ProgrammableMarketing/openworklabs.com (opens in a new tab)
PDF

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 8
  • This quarter · 11
  • Big bets · 3
  • Wildcards · 1
  • How we can help
Generated Nov 10, 2026
  1. Programmable
  2. Playbooks
  3. OpenWork
OpenWork logo

growth playbook · Nov 10, 2026

openworklabs.com (opens in a new tab)

OpenWork growth playbook

Diagnosis

OpenWork is a four-person Spring 2026 YC company with real pull: 24.5k GitHub stars, close to 200k downloads, weekly releases and a clear position as the open, any-model alternative to the big vendors' agent apps. The site is less ready than the product.

The homepage Windows button downloads the ARM64 build, there is no customer proof anywhere, the pricing page has no FAQ or annual option, sales is an unqualified calendar link, and the only way to hear about the weekly releases is to visit the changelog.

This playbook first fixes the download funnel, adds proof, pricing answers, a short sales form and a licence page. Next it builds the contributor and community engine, the desktop-to-Cloud upgrade moment and IT-buyer lead capture. The big bets are a packaged governed pilot, marketplace listings and a spend benchmark report built on gateway data.

Growth Score

6/10

Biggest Opportunity

Turn close to 200k anonymous desktop downloads into known teams, by fixing the download path, giving users a reason to share (templates, Cloud free) and a way to hear from OpenWork (release email, community).

Biggest Risk

With a team of four, a 331-PR backlog and well-funded rivals (the model vendors' own agent apps, plus goose and AnythingLLM with far larger contributor bases), OpenWork could fall behind on trust and polish even as its feature list keeps up.

01

Start here

The three moves to make first. Each one is written up in full below.

  1. 1

    Point the homepage Windows button at the x64 installer, not the ARM64 build

    Swap the homepage Windows href to openwork-win-x64-0.18.65.exe.

    High impact First result in 1 week
  2. 2

    Answer the pricing questions buyers ask: an FAQ, an annual Team price and the cost calculator

    Write the six FAQ answers.

    High impact First result in 2 weeks
  3. 3

    Turn the weekly release train into a weekly email

    Publish an RSS feed of the changelog.

    Medium impact First result in 1 month

02

Your idea portfolio

23 ideas, grouped by when to run them. Open any idea for how to run it: the steps, who to approach, copy you can use, and how to measure it.

  • 8 ideas are marked Unexpected: plays you are unlikely to have tried.

Every idea placed by the effort it takes against the impact it could have. Quick wins sit top left; select a number to open that idea. The filters below narrow the list by status, timing, category and what Programmable can build.

Effort vs impactEach number is an idea. Select one to open it.
Impact →
Quick wins · 2Big bets · 9Fill-ins · 8Reconsider · 4
Effort →

Showing 23 of 23 ideas

This week 8

Quick wins with a clear first action.

  • 01

    The homepage hero offers three platforms. macOS goes to /download, and Linux goes to the x86_64 AppImage, but the Windows link downloads openwork-win-arm64-0.18.65.exe, the ARM64 build. Most Windows PCs run x64 processors, and an ARM64 installer will not run on them. So the most common Windows visitor, clicking the most visible button on the site, gets a file that fails to install. The x64 installer already exists on /download; the homepage just does not link it. I checked the homepage twice on November 9, 2026 and found the same link both times.

    High impactLow effort First result in 1 week Founding Engineer

    First action

    Swap the homepage Windows href to openwork-win-x64-0.18.65.exe.

    How to measure it

    Windows first launches per week

    Now
    –
    Target
    +30%
    By
    30 days

    Steps

    1. 1

      Swap the link

      Point the homepage Windows button at the x64 installer.

      Founding Engineer 15 minutes

    2. 2

      Detect the platform

      Read OS and architecture in the browser; default to x64 on Windows when unsure.

      Founding Engineer 1 day

    3. 3

      Recommend one build

      On /download, highlight the detected build and move the rest under "Other builds".

      Design Engineer 2 days

    4. 4

      Generate links from releases

      Build homepage and /download links from the release manifest on every release.

      Founding Engineer 1 day

    5. 5

      Add a release check

      Fail the release if the default Windows link is not the x64 file.

      Founding Engineer 2 hours

    6. 6

      Track installs

      PostHog events for download by OS and architecture, and first launch.

      CTO 1 day

    7. 7

      Compare

      Windows first launches per download, four weeks before and after.

      Founder & CEO 4 weeks

    First result
    1 week
    Cost
    $0, about two days of engineering
    Owner
    Founding Engineer

    Smallest test

    Change the link today and compare Windows first launches over the next week.

    Why it works

    Today, change the homepage Windows link to the x64 installer. Then make the hero buttons architecture-aware: detect OS and architecture in the browser (Chromium browsers report architecture through User-Agent Client Hints), default to x64 when unsure, and add an "Other builds" link to /download. Give /download the same detection so it highlights one recommended build and tucks the others away. Finally, generate the homepage and /download links from the release manifest in the release workflow, so each weekly version updates them automatically and a check fails if the default Windows link points at an ARM64 file. Track download clicks and first launches by OS and architecture in PostHog, which the site already runs.

    For a free desktop app, the download button is the top of the funnel. A failed install never shows up as an error in web analytics; it looks like a visitor who simply did not convert. With ten releases in five weeks, a hand-edited link will drift again unless the links are generated. Windows laptops are also where most IT-managed seats sit, so these are the users most likely to become Team and Enterprise seats.

    Ready to use

    ChecklistDownload button check

    [ ] Homepage Windows button -> openwork-win-x64-<version>.exe
    [ ] Homepage macOS button -> detected Apple Silicon or Intel .dmg
    [ ] Homepage Linux button -> x86_64 AppImage (ARM64 offered when detected)
    [ ] "Other builds" link -> /download
    [ ] /download highlights one recommended build
    [ ] Links generated from the release manifest
    [ ] Release fails if the default Windows link is ARM64
    [ ] PostHog: download_clicked {os, arch}, first_launch {os, arch}

    Risks and how to handle them

    • Detection is wrong on some browsers

      Default to x64 and always show the other builds one click away.

    • Windows on ARM users get the x64 build

      Offer ARM64 when the browser reports it; Windows on ARM can also run x64 apps under emulation.

    Evidence

    • Your websiteHomepage (checked twice on November 9, 2026): H1 "Sovereign AI for knowledge workers" and subhead "Your AI workspace. Without vendor lock-in."; the hero platform links read macOS (to /download), Windows and Linux, but the Windows link downloads "openwork-win-arm64-0.18.65.exe", the ARM64 build, while the Linux link is the x86_64 AppImage; badges "Backed by YCombinator", "SOC 2 Type II" and 24.5K GitHub stars; "Free & open source · No account required to start"; "Talk to sales" opens a Google Calendar booking page; no customer logos, testimonials or newsletter signup.Source
    • Your websiteDownload page: separate Windows "x64 Installer" and "ARM64 Installer", macOS Apple Silicon and Intel builds and four Linux builds (v0.18.65); no automatic platform detection and no guidance on which build to choose beyond the label "Apple Silicon (M1+)".Source
    • Your websiteChangelog: ten releases from October 3 to November 5, 2026 (v0.18.56 to v0.18.65), including one-prompt migration of Claude Cowork and Code setups (v0.18.57), self-hosted OpenWork Web and team tools on AWS (v0.18.58), OneDrive and SharePoint files (v0.18.62), importing Claude chats into workspaces (v0.18.63) and Mac computer use across apps (v0.18.65); no RSS feed or email subscription.Source
  • 02

    OpenWork's own job post says the app has been downloaded close to 200k times, yet no page on the site uses that number. The homepage, enterprise and Cowork comparison pages show stars, the YC badge and SOC 2, but no named customer, logo or quote. IT and security buyers look for peers before they book a demo.

    High impactMedium effort First result in 1 month Founder & CEO

    First action

    Email the ten largest Cloud orgs asking for a 15-minute story call.

    How to measure it

    Demos booked per month

    Now
    –
    Target
    +25%
    By
    60 days

    Steps

    1. 1

      Add the number

      Downloads next to stars, from release stats.

      Founding Engineer 1 day

    2. 2

      Pick teams

      The ten largest Cloud orgs and self-hosted evaluators.

      Founder & CEO 1 day

    3. 3

      Interview

      Replaced what, seats, models, spend limits.

      Founder & CEO 1 week

    4. 4

      Write stories

      Three short stories, approved by each team.

      Founder & CEO 1 week

    5. 5

      Place proof

      Homepage, /enterprise, /pricing and the comparison page.

      Design Engineer 2 days

    6. 6

      Measure

      Demo bookings before and after.

      CTO monthly

    First result
    1 month
    Cost
    Founder time
    Owner
    Founder & CEO

    Smallest test

    Add the download count to the homepage this week.

    Why it works

    Put the download count (updated from release stats) next to the stars on the homepage and enterprise page. Ask Team and self-hosted users for three short stories, with permission: what they replaced, how many seats, which models and how they use spend limits. Add a logo row once two teams agree, and put one quote each on /pricing and the Cowork comparison page.

    Peer proof is what turns a good feature table into a booked demo, and the download number already exists.

    Risks and how to handle them

    • Teams won't be named

      Publish anonymised stories by role and industry until they agree.

    Evidence

    • HiringFounding Engineer (San Francisco, in person, $150K-$200K, 3-5% equity): works across the desktop app, agent runtime, browser automation, MCP and tool integrations, local permissions, cloud sync, realtime collaboration and plugin and skill infrastructure, shipping within 30 days and owning a major area within 60-90 days; the post says the app has been downloaded "close to 200k times".Source
    • Your websiteHomepage (checked twice on November 9, 2026): H1 "Sovereign AI for knowledge workers" and subhead "Your AI workspace. Without vendor lock-in."; the hero platform links read macOS (to /download), Windows and Linux, but the Windows link downloads "openwork-win-arm64-0.18.65.exe", the ARM64 build, while the Linux link is the x86_64 AppImage; badges "Backed by YCombinator", "SOC 2 Type II" and 24.5K GitHub stars; "Free & open source · No account required to start"; "Talk to sales" opens a Google Calendar booking page; no customer logos, testimonials or newsletter signup.Source
    • Your websiteEnterprise page: "AI your team wants. Control IT needs." for IT and security leaders: SSO, SCIM, per-team spend limits that warn at 90% and pause at 100%, providers such as AWS Bedrock, and self-hosting; model-gateway support for other coding agents is "still in progress"; every "Book a demo" goes to a calendar page; no named customers, logos or case studies.Source
  • 03

    /pricing has no FAQ, no annual Team price and no Team trial. Team and the managed-models add-on both cost $10 per user, and the Cloud Computer is priced per member, so buyers can easily misread what they are paying for. The most persuasive pricing asset, a 50-person, 3-year cost calculator, lives only on the Cowork comparison page. Product Hunt commenters asked directly what OpenWork costs to run.

    High impactLow effort First result in 2 weeks Founder & CEO

    First action

    Write the six FAQ answers.

    How to measure it

    Pricing-to-Team conversion

    Now
    –
    Target
    +20%
    By
    60 days

    Steps

    1. 1

      Write the FAQ

      Six questions, one paragraph each.

      Founder & CEO 1 day

    2. 2

      Fix units

      "Per user" everywhere, add-ons clearly marked.

      Design Engineer 2 hours

    3. 3

      Set annual price

      Annual Team option in billing.

      CTO 2 days

    4. 4

      Embed the calculator

      Reuse the comparison-page calculator on /pricing and /enterprise.

      Founding Engineer 1 day

    5. 5

      Add FAQ schema

      So search and AI assistants can quote the answers.

      Founding Engineer 1 hour

    6. 6

      Measure

      Checkout starts from /pricing.

      CTO monthly

    First result
    2 weeks
    Cost
    Staff time only
    Owner
    Founder & CEO

    Smallest test

    Publish the FAQ this week.

    Why it works

    Add a short FAQ covering the difference between Team and the OpenWork models add-on, what free self-hosting includes, what happens at the sixth user, and how billing works. Offer an annual Team price alongside monthly, use "per user" everywhere, and embed the cost calculator on /pricing and /enterprise.

    Clear totals remove the most common objection before a call, and an annual option improves cash and retention for a small team.

    Risks and how to handle them

    • Annual discount lowers revenue per seat

      Keep the discount small and test it on new teams only.

    Evidence

    • Your websitePricing page: Free at $0 for "up to 5 users"; Team at "$10 per user / month" with SSO/SAML, Extension Marketplace, cloud automations and "Basic usage analytics", "Billed monthly. Cancel anytime."; Enterprise at "Custom pricing" with SCIM, white-labeling and bring-your-own inference, "Annual contract with a 60-day opt-out"; add-ons "OpenWork Cloud Computer" at $50 per member a month and "OpenWork models" at $10 per user a month; no FAQ, no annual Team price and no Team trial; self-hosting can be evaluated free for 30 days at any size.Source
    • Your websiteClaude Cowork comparison page: "The free, open-source alternative to Claude Cowork"; a 25-row feature table (OpenWork 18 Yes plus 6 Partial, Claude Enterprise 19 Yes plus 2 Partial) checked against vendor docs on 2026-10-27; an illustrative 50-person, 3-year cost calculator (about $180,000 against about $72,900 on the same model); "No account needed"; no named customers or quotes.Source
    • Product Hunt59 points and "No reviews yet", launched in 2026; commenters asked whether it is "safe", what it costs to run compared with Claude Cowork, and whether it works with Slack or Notion.Source
  • 04

    Every "Talk to sales" and "Book a demo" button opens a Google Calendar booking page that asks nothing. OpenWork recently moved booking from Cal.com to a Google appointment schedule. With a team of four, every founder call with a three-person team that needed the self-serve plan is time not spent on a 200-seat rollout.

    Medium impactLow effort First result in 2 weeks Founder & CEO

    First action

    Write the three questions and the routing rule.

    How to measure it

    Qualified demo share

    Now
    –
    Target
    60%
    By
    60 days

    Steps

    1. 1

      Write questions

      Size, deployment, providers.

      Founder & CEO 1 hour

    2. 2

      Build the form

      One step before the calendar.

      Founding Engineer 1 day

    3. 3

      Route

      Small teams to self-serve, large teams to the founder.

      Founding Engineer 1 day

    4. 4

      Prep summary

      Answers sent with each booking.

      Founding Engineer 2 hours

    5. 5

      Self-serve email

      Guide and Cloud link for small teams.

      Founder & CEO 2 hours

    6. 6

      Review

      Booking volume and call quality monthly.

      Founder & CEO monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    Founder & CEO

    Smallest test

    Add the form to the enterprise page only.

    Why it works

    Put a three-question step before the calendar: team size, how they plan to deploy (desktop, Cloud or self-hosted) and which model providers they use. Teams under six get the free Cloud link and a setup guide; teams of 50 or more go to the founder's calendar. Every booking arrives with a prep summary.

    Short qualifying forms barely reduce bookings and make each call better prepared, which matters most when founders run every call.

    Risks and how to handle them

    • Fewer bookings

      Keep it to three questions and compare bookings for two weeks.

    Evidence

    • Your websiteHomepage (checked twice on November 9, 2026): H1 "Sovereign AI for knowledge workers" and subhead "Your AI workspace. Without vendor lock-in."; the hero platform links read macOS (to /download), Windows and Linux, but the Windows link downloads "openwork-win-arm64-0.18.65.exe", the ARM64 build, while the Linux link is the x86_64 AppImage; badges "Backed by YCombinator", "SOC 2 Type II" and 24.5K GitHub stars; "Free & open source · No account required to start"; "Talk to sales" opens a Google Calendar booking page; no customer logos, testimonials or newsletter signup.Source
    • Your websiteEnterprise page: "AI your team wants. Control IT needs." for IT and security leaders: SSO, SCIM, per-team spend limits that warn at 90% and pause at 100%, providers such as AWS Bedrock, and self-hosting; model-gateway support for other coding agents is "still in progress"; every "Book a demo" goes to a calendar page; no named customers, logos or case studies.Source
    • HiringDesign Engineer - SF ($130K-$200K, 1-5% equity) is also open; YC lists OpenWork as founded in 2026, Spring 2026 batch, a team of 4 in San Francisco, with LinkedIn and X (@getopenwork) profiles.Source
    • Your websiteSince April 2026 the homepage headline changed from "The open source Claude Cowork for your team" to "Your AI workspace. Without vendor lock-in.", the starting price from $50 a month to $10 per user, and sales booking from Cal.com to a Google appointment schedule; the page title still reads "OpenWork — Free, open-source Claude Cowork alternative".Source
  • 05

    The homepage now leads with "Sovereign AI for knowledge workers" and "Your AI workspace. Without vendor lock-in.", but the page title that search results show still reads "OpenWork — Free, open-source Claude Cowork alternative". Searchers see one pitch and visitors another. The alternative phrasing is what people actually search for, and the dedicated comparison page already targets it.

    Medium impactLow effort First result in 2 weeks Founder & CEO

    First action

    Add the open-source alternative line under the H1.

    How to measure it

    Search CTR

    Now
    –
    Target
    +10%
    By
    60 days

    Steps

    1. 1

      Pick the promise

      One line that matches the searches.

      Founder & CEO 1 hour

    2. 2

      Update the hero

      Show the line near the H1.

      Design Engineer 2 hours

    3. 3

      Retitle key pages

      /enterprise, /shadow-ai and /cloud.

      Founder & CEO 2 hours

    4. 4

      Check previews

      Social and chat link previews.

      Design Engineer 1 hour

    5. 5

      Watch Search Console

      CTR by query for four weeks.

      CTO 4 weeks

    6. 6

      Adjust

      One change at a time.

      Founder & CEO monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    Founder & CEO

    Smallest test

    Add one visible line under the H1 today.

    Why it works

    Keep the high-intent "open-source alternative" phrase in the homepage title, and repeat it visibly near the H1 so the click and the page match. Leave the vendor-specific keyword to /alternatives/claude-cowork. Give /enterprise and /shadow-ai their own governance-focused titles and descriptions.

    When the snippet and the page match, more searchers click and fewer bounce, and AI summaries describe the product the same way.

    Risks and how to handle them

    • Losing a ranking

      Keep the current keyword in the title and change only the page body first.

    Evidence

    • Your websiteSince April 2026 the homepage headline changed from "The open source Claude Cowork for your team" to "Your AI workspace. Without vendor lock-in.", the starting price from $50 a month to $10 per user, and sales booking from Cal.com to a Google appointment schedule; the page title still reads "OpenWork — Free, open-source Claude Cowork alternative".Source
    • Your websiteHomepage (checked twice on November 9, 2026): H1 "Sovereign AI for knowledge workers" and subhead "Your AI workspace. Without vendor lock-in."; the hero platform links read macOS (to /download), Windows and Linux, but the Windows link downloads "openwork-win-arm64-0.18.65.exe", the ARM64 build, while the Linux link is the x86_64 AppImage; badges "Backed by YCombinator", "SOC 2 Type II" and 24.5K GitHub stars; "Free & open source · No account required to start"; "Talk to sales" opens a Google Calendar booking page; no customer logos, testimonials or newsletter signup.Source
    • Your websiteClaude Cowork comparison page: "The free, open-source alternative to Claude Cowork"; a 25-row feature table (OpenWork 18 Yes plus 6 Partial, Claude Enterprise 19 Yes plus 2 Partial) checked against vendor docs on 2026-10-27; an illustrative 50-person, 3-year cost calculator (about $180,000 against about $72,900 on the same model); "No account needed"; no named customers or quotes.Source
  • 06

    OpenWork shipped ten releases in five weeks, including team-worthy features: migration in one prompt, SharePoint files, chat import and computer use. The changelog has no RSS feed or email signup, and the desktop app needs no account. OpenWork therefore cannot reach most of its close to 200k downloaders when it ships.

    Medium impactLow effort First result in 1 month Founding Engineer

    First action

    Publish an RSS feed of the changelog.

    How to measure it

    Release-digest subscribers

    Now
    –
    Target
    5,000
    By
    90 days

    Steps

    1. 1

      RSS feed

      Generated from the changelog.

      Founding Engineer 2 hours

    2. 2

      Signup

      On the changelog and /download.

      Founding Engineer 1 day

    3. 3

      In-app panel

      What's new, with an optional email field.

      Design Engineer 3 days

    4. 4

      First digest

      Top three releases with a try-this.

      Founder & CEO 2 hours

    5. 5

      Share link

      Invite-a-teammate link in every email.

      Founding Engineer 2 hours

    6. 6

      Measure

      Opens, returning launches, Cloud signups.

      CTO monthly

    First result
    1 month
    Cost
    Email tool, under $50 a month
    Owner
    Founding Engineer

    Smallest test

    Add RSS and a signup box to the changelog.

    Why it works

    Add an RSS feed and an email signup to the changelog, /download and an in-app "What's new" panel. Send a short monthly digest of the three biggest releases, each with a one-line "try this" and a link to share with a team. Offer an optional email field at first launch, never required.

    A release is the cheapest reason to bring a lapsed user back, and an opted-in email list is the first owned channel for a product that needs no account.

    Risks and how to handle them

    • Seen as tracking in a privacy-first product

      Keep it opt-in, separate from the app, with one-click unsubscribe.

    Evidence

    • Your websiteChangelog: ten releases from October 3 to November 5, 2026 (v0.18.56 to v0.18.65), including one-prompt migration of Claude Cowork and Code setups (v0.18.57), self-hosted OpenWork Web and team tools on AWS (v0.18.58), OneDrive and SharePoint files (v0.18.62), importing Claude chats into workspaces (v0.18.63) and Mac computer use across apps (v0.18.65); no RSS feed or email subscription.Source
    • Your websiteHomepage (checked twice on November 9, 2026): H1 "Sovereign AI for knowledge workers" and subhead "Your AI workspace. Without vendor lock-in."; the hero platform links read macOS (to /download), Windows and Linux, but the Windows link downloads "openwork-win-arm64-0.18.65.exe", the ARM64 build, while the Linux link is the x86_64 AppImage; badges "Backed by YCombinator", "SOC 2 Type II" and 24.5K GitHub stars; "Free & open source · No account required to start"; "Talk to sales" opens a Google Calendar booking page; no customer logos, testimonials or newsletter signup.Source
    • HiringFounding Engineer (San Francisco, in person, $150K-$200K, 3-5% equity): works across the desktop app, agent runtime, browser automation, MCP and tool integrations, local permissions, cloud sync, realtime collaboration and plugin and skill infrastructure, shipping within 30 days and owning a major area within 60-90 days; the post says the app has been downloaded "close to 200k times".Source
  • 07

    OpenWork sells itself as "Free & open source", but the code is under three licences: MIT outside ee/, the source-available OpenWork EE License inside ee/, and FSL-1.1-MIT for older versions. The pricing page adds more rules: free self-hosting up to 5 users without Enterprise features, plus a 30-day evaluation at any size. Developers and the legal teams of enterprise buyers both need a single clear map.

    Medium impactLow effort First result in 2 weeks Founder & CEO

    First action

    Draft the component-by-licence table from the README.

    How to measure it

    Legal-review cycle time

    Now
    –
    Target
    -30%
    By
    90 days

    Steps

    1. 1

      Map components

      Each package or folder and its licence.

      CTO 1 day

    2. 2

      Write FAQ

      Fork, self-host, evaluation, after the trial.

      Founder & CEO 1 day

    3. 3

      Legal check

      Counsel reviews the wording.

      Founder & CEO 3 days

    4. 4

      Publish

      /license, linked from the badge, pricing and README.

      Founding Engineer 2 hours

    5. 5

      Brief sales

      Send the page in every security review.

      Founder & CEO ongoing

    6. 6

      Review

      Update whenever a licence changes.

      CTO per change

    First result
    2 weeks
    Cost
    A few hours of counsel time
    Owner
    Founder & CEO

    Smallest test

    Add the table to the README first.

    Why it works

    Create a /license page with a table: component, licence, what you can do free, what needs a subscription. Add a short FAQ (can I fork it, can I self-host for 50 people, what happens after 30 days). Link it from the open-source badge, /pricing, /enterprise and the README.

    For an open-source alternative, licence clarity is part of the product promise; making it explicit removes a common blocker in security and legal review.

    Risks and how to handle them

    • Highlighting the source-available part

      Lead with what is MIT and free; be plain about the rest.

    Evidence

    • GitHubRepository different-ai/openwork: 24.5k stars, 2.5k forks, 82 open issues and 331 open pull requests; the licence is split: "Everything outside `ee/` is MIT", code in ee/ is under the source-available OpenWork EE License, and versions released before that remain FSL-1.1-MIT; the README links no Discord, forum or X account.Source
    • Your websitePricing page: Free at $0 for "up to 5 users"; Team at "$10 per user / month" with SSO/SAML, Extension Marketplace, cloud automations and "Basic usage analytics", "Billed monthly. Cancel anytime."; Enterprise at "Custom pricing" with SCIM, white-labeling and bring-your-own inference, "Annual contract with a 60-day opt-out"; add-ons "OpenWork Cloud Computer" at $50 per member a month and "OpenWork models" at $10 per user a month; no FAQ, no annual Team price and no Team trial; self-hosting can be evaluated free for 30 days at any size.Source
    • Your websiteHomepage (checked twice on November 9, 2026): H1 "Sovereign AI for knowledge workers" and subhead "Your AI workspace. Without vendor lock-in."; the hero platform links read macOS (to /download), Windows and Linux, but the Windows link downloads "openwork-win-arm64-0.18.65.exe", the ARM64 build, while the Linux link is the x86_64 AppImage; badges "Backed by YCombinator", "SOC 2 Type II" and 24.5K GitHub stars; "Free & open source · No account required to start"; "Talk to sales" opens a Google Calendar booking page; no customer logos, testimonials or newsletter signup.Source
  • 08

    "Openwork" is also the name of an unrelated staffing firm with an active Greenhouse job board, and OpenWork's own footer credits Different AI. YC lists OpenWork's LinkedIn and X profiles, but neither the site nor the README links them. Searchers, candidates and AI assistants get mixed signals about who OpenWork is.

    Medium impactLow effort First result in 1 month Founder & CEO

    First action

    Add footer links to X, LinkedIn and GitHub.

    How to measure it

    Clicks on "openwork" queries

    Now
    –
    Target
    +20%
    By
    90 days

    Steps

    1. 1

      Footer links

      X, LinkedIn, GitHub and YC.

      Design Engineer 1 hour

    2. 2

      Structured data

      Organization with sameAs and legal name.

      Founding Engineer 2 hours

    3. 3

      README links

      Site, social and community.

      CTO 1 hour

    4. 4

      One descriptor

      Same line on every profile.

      Founder & CEO 2 hours

    5. 5

      Check assistants

      Ask major assistants "what is OpenWork".

      Founder & CEO 1 hour

    6. 6

      Track

      Branded queries in Search Console.

      CTO monthly

    First result
    1 month
    Cost
    $0
    Owner
    Founder & CEO

    Smallest test

    Add footer social links and structured data today.

    Why it works

    Add Organization structured data (name, legal name Different AI, logo, and sameAs links to GitHub, YC, LinkedIn and X). Put social links in the site footer and the README. Use one consistent descriptor everywhere ("OpenWork, the open-source AI workspace"), and make sure job posts link back to openworklabs.com.

    Clear entity signals help search engines and assistants attach the brand to the right company, which protects branded search as awareness grows.

    Risks and how to handle them

    • Name confusion persists

      Use "OpenWork AI" or the full descriptor in titles where needed.

    Evidence

    • Brand researchThe name is shared: the Greenhouse job board at job-boards.greenhouse.io/openwork belongs to an unrelated staffing firm ("Modern Day Staffing, Reimagined") with 17 openings, mostly in Georgia and Texas; OpenWork's own footer reads "© 2026 Different AI".Source
    • HiringDesign Engineer - SF ($130K-$200K, 1-5% equity) is also open; YC lists OpenWork as founded in 2026, Spring 2026 batch, a team of 4 in San Francisco, with LinkedIn and X (@getopenwork) profiles.Source
    • GitHubRepository different-ai/openwork: 24.5k stars, 2.5k forks, 82 open issues and 331 open pull requests; the licence is split: "Everything outside `ee/` is MIT", code in ee/ is under the source-available OpenWork EE License, and versions released before that remain FSL-1.1-MIT; the README links no Discord, forum or X account.Source

This quarter 11

The core plays for the next three months.

  • 09

    The repository has 331 open pull requests against 82 open issues, so contributions are arriving four times faster than issues and waiting. goose has 600+ contributors and AnythingLLM 200+. With a team of four, unreviewed work is the fastest way to lose the people who would grow the project for free.

    High impactMedium effort First result in 1 month CTO

    First action

    Label all 331 PRs by area and size.

    How to measure it

    Median time to first review

    Now
    –
    Target
    Under 48 hours
    By
    90 days

    Steps

    1. 1

      Label

      Area, size, stale.

      CTO 2 days

    2. 2

      Triage sprint

      Merge small, close stale with thanks.

      CTO 2 weeks

    3. 3

      Contributing guide

      Which areas accept outside work.

      CTO 1 day

    4. 4

      First-reply bot

      Acknowledgement and expected time.

      Founding Engineer 1 day

    5. 5

      Weekly review hour

      Rotating reviewer.

      CTO weekly

    6. 6

      Good first issues

      A monthly list.

      Founding Engineer monthly

    7. 7

      Thank contributors

      Name them in release notes.

      Founder & CEO per release

    First result
    1 month
    Cost
    About a day a week of engineering time
    Owner
    CTO

    Smallest test

    Reply to the 20 newest PRs this week.

    Why it works

    Run a two-week triage sprint: label every PR, merge the small ones, close stale ones with a thank-you, and pin a CONTRIBUTING guide that says which areas accept outside work. Then set a 48-hour first-reply rule (a bot acknowledgement plus a human within two days), a weekly review hour and a monthly "good first issue" list. Once hired, the Founding Engineer owns plugin and skill contributions.

    Contributors who get a fast first review come back; each merged contributor is also a public advocate.

    Risks and how to handle them

    • Review time slows product work

      Cap it at a weekly hour and close out-of-scope PRs quickly.

    Evidence

    • GitHubRepository different-ai/openwork: 24.5k stars, 2.5k forks, 82 open issues and 331 open pull requests; the licence is split: "Everything outside `ee/` is MIT", code in ee/ is under the source-available OpenWork EE License, and versions released before that remain FSL-1.1-MIT; the README links no Discord, forum or X account.Source
    • Competitor researchAnythingLLM: "Own your intelligence.", a free MIT-licensed desktop app with "67k+" GitHub stars, 200+ contributors and a logo row; goose: "Your open source AI agent", maintained by the Agentic AI Foundation (Linux Foundation), with 55k+ stars, 600+ contributors, a Discord, an extensions directory, a community spotlight and shareable "Recipes".Source
    • HiringFounding Engineer (San Francisco, in person, $150K-$200K, 3-5% equity): works across the desktop app, agent runtime, browser automation, MCP and tool integrations, local permissions, cloud sync, realtime collaboration and plugin and skill infrastructure, shipping within 30 days and owning a major area within 60-90 days; the post says the app has been downloaded "close to 200k times".Source
  • 10

    Neither the site nor the README links a Discord, forum or X account, and the roadmap collects feedback only through a form. goose runs a Discord, a community spotlight and an extensions directory. On Product Hunt, users asked about safety, cost and Slack or Notion support in public comments, and those questions had nowhere to land.

    Medium impactMedium effort First result in 1 month Founder & CEO

    First action

    Turn on GitHub Discussions with four categories.

    How to measure it

    Weekly posters

    Now
    –
    Target
    150
    By
    90 days

    Steps

    1. 1

      Choose the space

      GitHub Discussions or Discord.

      Founder & CEO 1 day

    2. 2

      Seed channels

      Help, skills, show-and-tell, feedback.

      Founder & CEO 1 day

    3. 3

      Link everywhere

      App help menu, footer, README, /download.

      Design Engineer 1 day

    4. 4

      Monthly call

      Demo of the latest release.

      Founder & CEO monthly

    5. 5

      Docs from questions

      Top questions become docs.

      Founding Engineer monthly

    6. 6

      Spotlight

      Feature one user build a month.

      Founder & CEO monthly

    First result
    1 month
    Cost
    $0
    Owner
    Founder & CEO

    Smallest test

    Turn on Discussions and link it in the README.

    Why it works

    Pick one space (GitHub Discussions is enough to start, or Discord), seed it with channels for help, skills and templates, and show-and-tell, and link it from the app's help menu, the footer, /download and the README. Hold a monthly 30-minute community call to demo the latest release. Turn repeated questions into docs.

    A visible community answers questions faster than a team of four can, and it gives the product a public pulse that buyers notice.

    Risks and how to handle them

    • A quiet space looks worse than none

      Start with Discussions and commit to same-day replies.

    Evidence

    • GitHubRepository different-ai/openwork: 24.5k stars, 2.5k forks, 82 open issues and 331 open pull requests; the licence is split: "Everything outside `ee/` is MIT", code in ee/ is under the source-available OpenWork EE License, and versions released before that remain FSL-1.1-MIT; the README links no Discord, forum or X account.Source
    • Competitor researchAnythingLLM: "Own your intelligence.", a free MIT-licensed desktop app with "67k+" GitHub stars, 200+ contributors and a logo row; goose: "Your open source AI agent", maintained by the Agentic AI Foundation (Linux Foundation), with 55k+ stars, 600+ contributors, a Discord, an extensions directory, a community spotlight and shareable "Recipes".Source
    • Product Hunt59 points and "No reviews yet", launched in 2026; commenters asked whether it is "safe", what it costs to run compared with Claude Cowork, and whether it works with Slack or Notion.Source
    • Your websiteRoadmap (updated October 27): shipped since August are AI Gateway spend limits by team, desktop policies, OpenWork Web, automations, audit logs and OpenWork in Slack; planned are a phone app, admin roles, the AI Gateway from other coding agents, approvals mid-run and failure and spend reporting by skill, connection and team; there is no voting, only a "Share feedback" form.Source
  • 11

    The desktop app needs no account, which is great for adoption and leaves OpenWork blind to who its users are. Cloud is free for up to 5 users and offers "Shareable workspace templates" that bundle skills and MCPs into a link. The natural moment to create an account is when someone wants a teammate to have the same setup.

    High impactMedium effort First result in 6 weeks Design Engineer

    First action

    Add a share button to the skills list that opens Cloud sign-up with the skill attached.

    How to measure it

    Install-to-org rate

    Now
    –
    Target
    8%
    By
    1 quarter

    Steps

    1. 1

      Design the moment

      Share button on skills, MCPs and workspaces.

      Design Engineer 1 week

    2. 2

      One-step org

      Create a free org with the item attached.

      Founding Engineer 2 weeks

    3. 3

      Invite link

      Recipient installs with the setup preloaded.

      Founding Engineer 1 week

    4. 4

      Fifth-member nudge

      Show what Team adds.

      Design Engineer 3 days

    5. 5

      Instrument

      Share, join and upgrade events.

      CTO 2 days

    6. 6

      Iterate

      Test copy and placement.

      Design Engineer monthly

    First result
    6 weeks
    Cost
    About three weeks of engineering
    Owner
    Design Engineer

    Smallest test

    Add a share link to one screen and count clicks.

    Why it works

    Add a "Share with my team" action next to every skill, MCP and workspace in the desktop app. It creates a free Cloud org in one step and produces an invite link. At the fifth member, show what Team adds (unlimited users, SSO, cloud automations) with a one-click upgrade. Measure how many desktop installs reach a shared org.

    The upgrade trigger for a free desktop agent is collaboration, not usage limits, and every share brings in a new user by invitation.

    Risks and how to handle them

    • Users resent account prompts

      Only show it when they choose to share; local use stays account-free.

    Evidence

    • Your websiteHomepage (checked twice on November 9, 2026): H1 "Sovereign AI for knowledge workers" and subhead "Your AI workspace. Without vendor lock-in."; the hero platform links read macOS (to /download), Windows and Linux, but the Windows link downloads "openwork-win-arm64-0.18.65.exe", the ARM64 build, while the Linux link is the x86_64 AppImage; badges "Backed by YCombinator", "SOC 2 Type II" and 24.5K GitHub stars; "Free & open source · No account required to start"; "Talk to sales" opens a Google Calendar booking page; no customer logos, testimonials or newsletter signup.Source
    • Your websitePricing page: Free at $0 for "up to 5 users"; Team at "$10 per user / month" with SSO/SAML, Extension Marketplace, cloud automations and "Basic usage analytics", "Billed monthly. Cancel anytime."; Enterprise at "Custom pricing" with SCIM, white-labeling and bring-your-own inference, "Annual contract with a 60-day opt-out"; add-ons "OpenWork Cloud Computer" at $50 per member a month and "OpenWork models" at $10 per user a month; no FAQ, no annual Team price and no Team trial; self-hosting can be evaluated free for 30 days at any size.Source
    • Your websiteCloud page: "OpenWork Cloud — the dashboard for your whole org": org-wide skills and MCP servers, central model provisioning, role-based access, usage and audit views, and "Shareable workspace templates" that bundle skills and MCPs into a link; the free plan covers up to 5 users.Source
  • 12

    The Shadow AI page links an AI governance guide, an AI usage policy builder and a rollout checklist, all aimed at IT, security and compliance leaders, and none of them asks for an email. These are exactly the buyers of Enterprise, and right now they read and leave.

    High impactMedium effort First result in 1 month Founder & CEO

    First action

    Export the policy builder and checklist into one editable document.

    How to measure it

    Kit downloads with work email

    Now
    –
    Target
    200 a month
    By
    90 days

    Steps

    1. 1

      Build the kit

      Policy, checklist and spend plan.

      Founder & CEO 1 week

    2. 2

      Add capture

      Work-email form for the editable version.

      Founding Engineer 1 day

    3. 3

      Write follow-ups

      Three emails ending in a pilot offer.

      Founder & CEO 2 days

    4. 4

      Promote

      LinkedIn posts and IT communities.

      Founder & CEO ongoing

    5. 5

      Route

      Large companies to the founder's calendar.

      Founding Engineer 1 day

    6. 6

      Measure

      Downloads to demos.

      CTO monthly

    First result
    1 month
    Cost
    Staff time and an email tool
    Owner
    Founder & CEO

    Smallest test

    Offer the checklist as a PDF for an email on /shadow-ai.

    Why it works

    Package the policy builder output, the rollout checklist and a one-page spend-limit plan into an editable kit (doc and PDF). Keep the web versions open, and offer the editable kit in return for a work email. Follow up with one useful email a week for three weeks (policy, rollout, spend) ending in a pilot offer.

    IT leaders need a policy before they need a tool; owning that first step puts OpenWork on the shortlist.

    Risks and how to handle them

    • Gating hurts SEO

      Keep the web pages open; gate only the editable file.

    Evidence

    • Your websiteShadow AI page: "Shadow AI governance. Keep work moving." for IT, security and compliance leaders, linking an AI governance guide, an AI usage policy builder and an on-page rollout checklist, none of which asks for an email address.Source
    • Your websiteEnterprise page: "AI your team wants. Control IT needs." for IT and security leaders: SSO, SCIM, per-team spend limits that warn at 90% and pause at 100%, providers such as AWS Bedrock, and self-hosting; model-gateway support for other coding agents is "still in progress"; every "Book a demo" goes to a calendar page; no named customers, logos or case studies.Source
  • 13

    The site's only comparison pages are the two Claude Cowork pages; there are none against other open-source agents such as goose (55k+ stars) or AnythingLLM (67k+), which developers evaluate side by side. The existing comparison page already shows the format: a cited feature table and a cost calculator.

    Medium impactMedium effort First result in 6 weeks Founder & CEO

    First action

    Draft the goose comparison table.

    How to measure it

    Comparison-page downloads

    Now
    –
    Target
    1,000 a month
    By
    90 days

    Steps

    1. 1

      Pick tools

      goose and AnythingLLM first.

      Founder & CEO 1 day

    2. 2

      Check facts

      From each tool's own docs, dated.

      Founding Engineer 2 days

    3. 3

      Write pages

      Same table format as the existing page.

      Founder & CEO 1 week

    4. 4

      Hub page

      /alternatives listing every comparison.

      Design Engineer 1 day

    5. 5

      Publish

      With sitemap entries.

      Founding Engineer 1 hour

    6. 6

      Refresh

      Quarterly fact check.

      Founder & CEO quarterly

    First result
    6 weeks
    Cost
    Staff time only
    Owner
    Founder & CEO

    Smallest test

    Publish the goose page and track downloads.

    Why it works

    Publish fair, dated comparison pages against goose and AnythingLLM, using the same table format and citing each tool's own docs. Lead with OpenWork's differentiators: team sharing of skills and MCPs, one MCP URL for every agent, spend limits and a self-hostable control plane. Add a hub page that lists every comparison.

    Comparison searches are high intent, and fair tables also teach AI assistants how the tools differ.

    Risks and how to handle them

    • Claims go out of date

      Date each page and recheck quarterly.

    Evidence

    • Your websiteSitemap (22 URLs, checked November 9, 2026): the homepage, two /alternatives pages (claude-cowork and claude-cowork-3p), four guides, /glm-5.2, /connect, /cloud, /dashboard, /download, /enterprise, /shadow-ai, /ai-governance, /pricing, /roadmap, /privacy and /terms; no blog, no customer stories and no comparison with any tool other than Claude Cowork; the docs (about 810 pages) are the only other content.Source
    • Competitor researchAnythingLLM: "Own your intelligence.", a free MIT-licensed desktop app with "67k+" GitHub stars, 200+ contributors and a logo row; goose: "Your open source AI agent", maintained by the Agentic AI Foundation (Linux Foundation), with 55k+ stars, 600+ contributors, a Discord, an extensions directory, a community spotlight and shareable "Recipes".Source
    • Your websiteClaude Cowork comparison page: "The free, open-source alternative to Claude Cowork"; a 25-row feature table (OpenWork 18 Yes plus 6 Partial, Claude Enterprise 19 Yes plus 2 Partial) checked against vendor docs on 2026-10-27; an illustrative 50-person, 3-year cost calculator (about $180,000 against about $72,900 on the same model); "No account needed"; no named customers or quotes.Source
  • 14

    In October OpenWork shipped one-prompt migration of Claude Cowork and Code setups (v0.18.57) and import of Claude chats into workspaces (v0.18.63). The comparison page already estimates large savings for a 50-person team. Yet the migration lives in a changelog line and a docs link, not in a campaign aimed at the teams who would switch.

    High impactMedium effort First result in 1 month Founder & CEO

    First action

    Record a 10-minute migration video.

    How to measure it

    Migrations per month

    Now
    –
    Target
    300
    By
    1 quarter

    Steps

    1. 1

      Record the video

      A real migration, start to finish.

      Founder & CEO 1 day

    2. 2

      Build /switch

      What moves, what changes, calculator.

      Design Engineer 1 week

    3. 3

      Track migrations

      An event when migration completes.

      Founding Engineer 1 day

    4. 4

      Promote

      LinkedIn and community posts.

      Founder & CEO weekly

    5. 5

      Office hours

      Live switcher help every two weeks.

      Founder & CEO biweekly

    6. 6

      Follow up

      Invite migrated users to share with their team.

      Founder & CEO ongoing

    First result
    1 month
    Cost
    Staff time; optional small ad budget
    Owner
    Founder & CEO

    Smallest test

    Post the migration video and count migrations that week.

    Why it works

    Build a /switch landing page: what moves (skills, plugins, chats), what changes, a 10-minute video of a real migration, and the cost calculator. Target team admins already on the incumbent tool through LinkedIn posts, community threads and a switcher office hour every two weeks. Track migrations completed as the success event.

    Switching cost is the main barrier for teams already set up elsewhere, and OpenWork has removed it; the campaign just has to say so.

    Risks and how to handle them

    • Migration fails for some setups

      Log failures and fix the top three before promoting widely.

    Evidence

    • Your websiteChangelog: ten releases from October 3 to November 5, 2026 (v0.18.56 to v0.18.65), including one-prompt migration of Claude Cowork and Code setups (v0.18.57), self-hosted OpenWork Web and team tools on AWS (v0.18.58), OneDrive and SharePoint files (v0.18.62), importing Claude chats into workspaces (v0.18.63) and Mac computer use across apps (v0.18.65); no RSS feed or email subscription.Source
    • Your websiteClaude Cowork comparison page: "The free, open-source alternative to Claude Cowork"; a 25-row feature table (OpenWork 18 Yes plus 6 Partial, Claude Enterprise 19 Yes plus 2 Partial) checked against vendor docs on 2026-10-27; an illustrative 50-person, 3-year cost calculator (about $180,000 against about $72,900 on the same model); "No account needed"; no named customers or quotes.Source
  • 15

    OpenWork Cloud can already bundle skills and MCPs into a shareable workspace template link, and the MCP Gateway works in seven agent clients. goose has an extensions directory and shareable recipes. OpenWork has no public page where those templates live, so every good setup stays private.

    High impactMedium effort First result in 6 weeks Design Engineer

    First action

    Publish ten templates the team already uses internally.

    How to measure it

    Template-driven installs

    Now
    –
    Target
    2,000 a month
    By
    2 quarters

    Steps

    1. 1

      Pick ten jobs

      The most common uses.

      Founder & CEO 1 day

    2. 2

      Build templates

      Skills and MCPs bundled.

      Founding Engineer 1 week

    3. 3

      Design pages

      One indexable page per template.

      Design Engineer 1 week

    4. 4

      One-click use

      Open in desktop or Cloud.

      Founding Engineer 1 week

    5. 5

      Submissions

      Community templates with credit.

      Design Engineer 2 weeks

    6. 6

      Promote

      One template a week on social.

      Founder & CEO weekly

    First result
    6 weeks
    Cost
    About a month of engineering and design
    Owner
    Design Engineer

    Smallest test

    Publish three template pages and track installs.

    Why it works

    Launch /templates: a searchable gallery of workspace templates by job (weekly team update from project tools and Drive, sales call prep, support triage, finance close). Each template gets its own indexable page with what it does, which connections it needs and a one-click "Use this template" that installs or opens Cloud. Let users submit their own, credited.

    Each template is a landing page for a specific search and an invite when someone shares it, so the gallery grows both search traffic and word of mouth.

    Risks and how to handle them

    • Low-quality submissions

      Review before listing and badge verified templates.

    Builds on

    Evidence

    • Your websiteCloud page: "OpenWork Cloud — the dashboard for your whole org": org-wide skills and MCP servers, central model provisioning, role-based access, usage and audit views, and "Shareable workspace templates" that bundle skills and MCPs into a link; the free plan covers up to 5 users.Source
    • Your websiteMCP Gateway page: "Add one URL. Use your MCPs in every app." for Claude Code, Codex, Cursor, VS Code, OpenCode, Gemini CLI and ChatGPT; Solo free, Team $10 per user a month.Source
    • Competitor researchAnythingLLM: "Own your intelligence.", a free MIT-licensed desktop app with "67k+" GitHub stars, 200+ contributors and a logo row; goose: "Your open source AI agent", maintained by the Agentic AI Foundation (Linux Foundation), with 55k+ stars, 600+ contributors, a Discord, an extensions directory, a community spotlight and shareable "Recipes".Source
    • Your websiteSitemap (22 URLs, checked November 9, 2026): the homepage, two /alternatives pages (claude-cowork and claude-cowork-3p), four guides, /glm-5.2, /connect, /cloud, /dashboard, /download, /enterprise, /shadow-ai, /ai-governance, /pricing, /roadmap, /privacy and /terms; no blog, no customer stories and no comparison with any tool other than Claude Cowork; the docs (about 810 pages) are the only other content.Source
  • 16

    The first question on OpenWork's Product Hunt page was whether it is safe; the maker answered that it only works in the folders you give it. A new user also has to choose a model and keys before anything happens, even though a guide shows how to run agents locally with Ollama and no API key. The Design Engineer and Founding Engineer roles (local permissions) are being hired for exactly this.

    High impactMedium effort First result in 6 weeks Design Engineer

    First action

    Measure today's share of installs that complete a first task.

    How to measure it

    Day-one activation

    Now
    –
    Target
    50%
    By
    90 days

    Steps

    1. 1

      Measure baseline

      First-task completion today.

      CTO 2 days

    2. 2

      Safety screen

      Folders, ask-before-actions, data flow.

      Design Engineer 1 week

    3. 3

      Model choice

      Local, own key or managed, in one step.

      Design Engineer 1 week

    4. 4

      Starter task

      Guided task on a sample folder.

      Founding Engineer 1 week

    5. 5

      Test

      Five user sessions, then ship.

      Design Engineer 1 week

    6. 6

      Track

      Activation by platform and model choice.

      CTO weekly

    First result
    6 weeks
    Cost
    About a month of design and engineering
    Owner
    Design Engineer

    Smallest test

    Add the safety screen alone and compare first-task completion.

    Why it works

    Redesign the first five minutes: a one-screen explanation of what the agent can and cannot touch (chosen folders, ask before actions), a choice between a local model with no key, your own key or managed models, and one guided starter task on a sample folder that completes in under two minutes. Measure the share of installs that finish a first task.

    Trust and a fast first success are what turn an install into a habit, and both are design problems more than feature gaps.

    Risks and how to handle them

    • Local models are slow on older laptops

      Recommend managed or own-key when hardware is limited.

    Evidence

    • Product Hunt59 points and "No reviews yet", launched in 2026; commenters asked whether it is "safe", what it costs to run compared with Claude Cowork, and whether it works with Slack or Notion.Source
    • Your websiteGuides: four pages covering OpenCode, other coding agents via one MCP URL, Ollama (local agents on your own files with no API key) and "How do you keep AI agent costs under control?".Source
    • HiringDesign Engineer - SF ($130K-$200K, 1-5% equity) is also open; YC lists OpenWork as founded in 2026, Spring 2026 batch, a team of 4 in San Francisco, with LinkedIn and X (@getopenwork) profiles.Source
    • HiringFounding Engineer (San Francisco, in person, $150K-$200K, 3-5% equity): works across the desktop app, agent runtime, browser automation, MCP and tool integrations, local permissions, cloud sync, realtime collaboration and plugin and skill infrastructure, shipping within 30 days and owning a major area within 60-90 days; the post says the app has been downloaded "close to 200k times".Source
  • 17

    Real visitors already pass Core Web Vitals (LCP 1.3 s, INP 121 ms), so this is not urgent. But in the lab, desktop scores 69, with 1.8 seconds of total blocking time, and PageSpeed flags short cache lifetimes (about 0.6 s) and 136 KB of unused JavaScript. Developers and IT evaluators mostly arrive on desktop.

    Low impactLow effort First result in 1 week Founding Engineer

    First action

    Lazy-load the app demo section.

    How to measure it

    Desktop performance score

    Now
    –
    Target
    90
    By
    30 days

    Steps

    1. 1

      Cache headers

      Long lifetimes on hashed assets.

      Founding Engineer 2 hours

    2. 2

      Lazy-load demo

      App demo and prompt widget on scroll.

      Founding Engineer 1 day

    3. 3

      Cut unused JS

      Bundle analysis on the homepage.

      Founding Engineer 1 day

    4. 4

      Fix contrast

      The flagged colour pair.

      Design Engineer 1 hour

    5. 5

      Retest

      PageSpeed desktop and mobile.

      Founding Engineer 1 hour

    6. 6

      Guard

      Performance budget in CI.

      CTO 1 day

    First result
    1 week
    Cost
    $0
    Owner
    Founding Engineer

    Smallest test

    Fix cache headers and retest.

    Why it works

    Set long cache lifetimes on hashed static assets, load the interactive app demo and the copy-prompt widget only when they scroll into view, and drop unused JavaScript from the homepage bundle. Fix the one colour-contrast issue at the same time.

    A snappy page fits a product sold on speed and polish, and the fixes are routine in Next.js.

    Risks and how to handle them

    • Demo appears late

      Reserve its space and show a static frame first.

    Evidence

    • Google PageSpeedTested November 10, 2026: mobile score 90 and desktop 69 (desktop total blocking time 1,808 ms); lab mobile LCP 2.8 s; real visitors pass Core Web Vitals (LCP 1.3 s, INP 121 ms, CLS 0); the biggest fixes are cache lifetimes (about 0.6 s) and 136 KB of unused JavaScript; one colour-contrast issue.Source
  • 18

    The roadmap is public and updated (October 27), with a long "Coming soon" list including a phone app, admin roles and mid-run approvals, but feedback only goes into a form with no voting and no follow-up. The changelog ships weekly, so many requests are fulfilled without the requester ever hearing about it.

    Medium impactLow effort First result in 1 month Founder & CEO

    First action

    Add a vote button with an email field to the five most-requested items.

    How to measure it

    Voters active within 7 days of ship

    Now
    –
    Target
    40%
    By
    90 days

    Steps

    1. 1

      Add votes

      Email-backed votes on roadmap items.

      Founding Engineer 2 days

    2. 2

      Tag requests

      Link form feedback to items.

      Founder & CEO 1 day

    3. 3

      Ship notes

      Email voters when an item ships.

      Founding Engineer 1 day

    4. 4

      Link releases

      Shipped items point to the changelog.

      Design Engineer 1 day

    5. 5

      Prioritise

      Use votes as one input.

      CTO monthly

    6. 6

      Measure

      Voter return rate.

      CTO monthly

    First result
    1 month
    Cost
    $0
    Owner
    Founder & CEO

    Smallest test

    Email everyone who asked for one feature when it ships.

    Why it works

    Add votes to roadmap items (each vote asks for an email). Tag each item with the requests behind it, and when it ships, email those voters with a one-line how-to. Mark shipped items with their release version and link the changelog entry.

    Hearing "you asked, we shipped" is one of the strongest retention moments for an open-source product, and it builds the email list.

    Risks and how to handle them

    • Votes skew to loud users

      Weigh votes alongside Team and Enterprise needs.

    Builds on

    Evidence

    • Your websiteRoadmap (updated October 27): shipped since August are AI Gateway spend limits by team, desktop policies, OpenWork Web, automations, audit logs and OpenWork in Slack; planned are a phone app, admin roles, the AI Gateway from other coding agents, approvals mid-run and failure and spend reporting by skill, connection and team; there is no voting, only a "Share feedback" form.Source
    • Your websiteChangelog: ten releases from October 3 to November 5, 2026 (v0.18.56 to v0.18.65), including one-prompt migration of Claude Cowork and Code setups (v0.18.57), self-hosted OpenWork Web and team tools on AWS (v0.18.58), OneDrive and SharePoint files (v0.18.62), importing Claude chats into workspaces (v0.18.63) and Mac computer use across apps (v0.18.65); no RSS feed or email subscription.Source
  • 19

    OpenWork ran a launch page for a new open model with a 2x usage boost on OpenWork Models, its $10-per-user managed inference add-on. That is a recurring opportunity: open models ship often, and each launch is a reason for users who bring their own keys to try the add-on. Today it is a one-off page.

    Medium impactLow effort First result in 1 month Founder & CEO

    First action

    Turn the existing launch page into a reusable template.

    How to measure it

    Team seats with the add-on

    Now
    –
    Target
    20%
    By
    2 quarters

    Steps

    1. 1

      Template

      Reusable launch page.

      Design Engineer 2 days

    2. 2

      Benchmark kit

      Five real OpenWork tasks to run on each model.

      Founding Engineer 1 week

    3. 3

      In-app banner

      Shown to own-key users.

      Founding Engineer 2 days

    4. 4

      Launch playbook

      48-hour checklist.

      Founder & CEO 1 day

    5. 5

      Run

      Next notable release.

      Founder & CEO per launch

    6. 6

      Measure

      Attach rate per launch.

      CTO per launch

    First result
    1 month
    Cost
    Inference cost of usage boosts
    Owner
    Founder & CEO

    Smallest test

    Add an in-app banner for the current launch.

    Why it works

    Turn the launch page into a template. For each notable open-model release, publish a page within 48 hours with a short benchmark on real OpenWork tasks, a time-limited usage boost and an in-app banner for users on their own keys. Track add-on attach rate per launch.

    Launch moments bring searches and attention; pairing them with a no-keys trial lifts add-on revenue without discounting Team.

    Risks and how to handle them

    • Usage boosts cost money

      Time-limit them and cap usage per seat.

    Evidence

    • Your websiteModel launch page: "GLM 5.2 is now in OpenWork — with 2x usage", promoting OpenWork Models, managed access to third-party open models at $10 per user a month, with repeated sign-up CTAs.Source
    • Your websitePricing page: Free at $0 for "up to 5 users"; Team at "$10 per user / month" with SSO/SAML, Extension Marketplace, cloud automations and "Basic usage analytics", "Billed monthly. Cancel anytime."; Enterprise at "Custom pricing" with SCIM, white-labeling and bring-your-own inference, "Annual contract with a 60-day opt-out"; add-ons "OpenWork Cloud Computer" at $50 per member a month and "OpenWork models" at $10 per user a month; no FAQ, no annual Team price and no Team trial; self-hosting can be evaluated free for 30 days at any size.Source

Big bets 3

Larger plays that can change the trajectory.

  • 20

    Enterprise already has a 60-day opt-out, and self-hosting can be evaluated free for 30 days at any size, but nothing on the site describes what a pilot looks like or how success is judged. With no named customers yet, the first three enterprise wins matter more than any page.

    High impactHigh effort First result in 2 months Founder & CEO

    First action

    Write the one-page pilot plan with its success criteria.

    How to measure it

    Pilot-to-contract rate

    Now
    –
    Target
    60%
    By
    2 quarters

    Steps

    1. 1

      Define the pilot

      Scope, setup and success criteria.

      Founder & CEO 1 week

    2. 2

      Build the report

      Weekly adoption and spend report.

      CTO 2 weeks

    3. 3

      Recruit partners

      Three companies from inbound and the IT kit.

      Founder & CEO 1 month

    4. 4

      Run pilots

      With a weekly check-in.

      Founder & CEO 30 days each

    5. 5

      Readout

      Results against the criteria.

      Founder & CEO 1 day each

    6. 6

      Publish

      The pilot page and case studies.

      Design Engineer 2 weeks

    First result
    2 months
    Cost
    Founder time and a first-year discount
    Owner
    Founder & CEO

    Smallest test

    Offer the pilot to the next inbound enterprise lead.

    Why it works

    Define a fixed 30-day pilot: one department of 25 to 100 people, SSO, two approved providers, spend limits, three shared templates and a weekly usage report. It ends with a readout on adoption, spend against budget and shadow-AI reduction. Recruit three design partners at a reduced first-year price in exchange for a case study, and publish the pilot as a page.

    A clear pilot shortens the enterprise cycle for a small team, and the design partners become the proof idea 2 needs.

    Risks and how to handle them

    • Pilots drift without an owner

      Require a named sponsor and fixed success criteria before starting.

    Builds on

    Evidence

    • Your websitePricing page: Free at $0 for "up to 5 users"; Team at "$10 per user / month" with SSO/SAML, Extension Marketplace, cloud automations and "Basic usage analytics", "Billed monthly. Cancel anytime."; Enterprise at "Custom pricing" with SCIM, white-labeling and bring-your-own inference, "Annual contract with a 60-day opt-out"; add-ons "OpenWork Cloud Computer" at $50 per member a month and "OpenWork models" at $10 per user a month; no FAQ, no annual Team price and no Team trial; self-hosting can be evaluated free for 30 days at any size.Source
    • Your websiteEnterprise page: "AI your team wants. Control IT needs." for IT and security leaders: SSO, SCIM, per-team spend limits that warn at 90% and pause at 100%, providers such as AWS Bedrock, and self-hosting; model-gateway support for other coding agents is "still in progress"; every "Book a demo" goes to a calendar page; no named customers, logos or case studies.Source
  • 21

    OpenWork already supports self-hosting OpenWork Web and its team tools on AWS (v0.18.58) and connects to AWS Bedrock as a provider. Enterprise is sold on annual contracts. Many IT buyers prefer to buy through their existing cloud account, which lets them draw down committed spend and skip a new vendor setup.

    High impactHigh effort First result in 1 quarter Founder & CEO

    First action

    List the requirements for a self-hosted marketplace listing.

    How to measure it

    Marketplace-sourced contracts

    Now
    –
    Target
    3
    By
    3 quarters

    Steps

    1. 1

      Scope

      Listing requirements and pricing model.

      Founder & CEO 1 week

    2. 2

      Package

      Deployment template for self-hosting.

      CTO 1 month

    3. 3

      Security docs

      Reuse the Trust Center material.

      CTO 1 week

    4. 4

      List

      Submit and publish the listing.

      Founder & CEO 1 month

    5. 5

      Link

      From /enterprise and the pilot page.

      Design Engineer 1 day

    6. 6

      Co-sell

      Engage the cloud partner team.

      Founder & CEO ongoing

    First result
    1 quarter
    Cost
    Marketplace fees on transactions plus engineering time
    Owner
    Founder & CEO

    Smallest test

    Ask pilot prospects whether they would buy through their cloud account.

    Why it works

    Package the self-hosted edition as a marketplace listing, starting with AWS, with Team and Enterprise pricing through the marketplace and a deployment template. Add the listing link to /enterprise and the pilot page. Then evaluate the other major cloud marketplaces based on where pilots run.

    Buying through an existing cloud account removes procurement steps, which matters most for a young vendor without a track record.

    Risks and how to handle them

    • Effort for a four-person team

      Start only after the first pilot asks to buy through its cloud account.

    Builds on

    Evidence

    • Your websiteChangelog: ten releases from October 3 to November 5, 2026 (v0.18.56 to v0.18.65), including one-prompt migration of Claude Cowork and Code setups (v0.18.57), self-hosted OpenWork Web and team tools on AWS (v0.18.58), OneDrive and SharePoint files (v0.18.62), importing Claude chats into workspaces (v0.18.63) and Mac computer use across apps (v0.18.65); no RSS feed or email subscription.Source
    • Your websiteEnterprise page: "AI your team wants. Control IT needs." for IT and security leaders: SSO, SCIM, per-team spend limits that warn at 90% and pause at 100%, providers such as AWS Bedrock, and self-hosting; model-gateway support for other coding agents is "still in progress"; every "Book a demo" goes to a calendar page; no named customers, logos or case studies.Source
    • Your websitePricing page: Free at $0 for "up to 5 users"; Team at "$10 per user / month" with SSO/SAML, Extension Marketplace, cloud automations and "Basic usage analytics", "Billed monthly. Cancel anytime."; Enterprise at "Custom pricing" with SCIM, white-labeling and bring-your-own inference, "Annual contract with a 60-day opt-out"; add-ons "OpenWork Cloud Computer" at $50 per member a month and "OpenWork models" at $10 per user a month; no FAQ, no annual Team price and no Team trial; self-hosting can be evaluated free for 30 days at any size.Source
  • 22

    OpenWork's AI Gateway already tracks spend, limits and usage by team, with warnings at 90% and pauses at 100%. Spend reporting by skill and connection is on the roadmap, and one of the four guides is about controlling AI agent costs. Few vendors can see how real teams split work across models and what it costs, and OpenWork's gateway can.

    Medium impactHigh effort First result in 2 quarters Founder & CEO

    First action

    Add an opt-in for anonymised benchmarking to the Cloud admin settings.

    How to measure it

    Leads from the report

    Now
    –
    Target
    300
    By
    first edition + 60 days

    Steps

    1. 1

      Opt-in

      Admin setting for anonymised benchmarking.

      CTO 1 week

    2. 2

      Privacy rules

      Aggregation and minimum group sizes.

      CTO 1 week

    3. 3

      Analyse

      Spend, model mix and limit events.

      CTO 2 weeks

    4. 4

      Write

      Findings and what efficient teams do.

      Founder & CEO 2 weeks

    5. 5

      Benchmark tool

      Admins compare their own numbers.

      Founding Engineer 2 weeks

    6. 6

      Launch

      Press, LinkedIn and IT communities.

      Founder & CEO 1 week

    First result
    2 quarters
    Cost
    About six weeks of team time
    Owner
    Founder & CEO

    Smallest test

    Publish five anonymised charts from opted-in orgs.

    Why it works

    With explicit opt-in and aggregation (no customer names, minimum group sizes), publish a report twice a year: spend per seat by team type, the mix of open and closed models, how often limits trigger and what the cheapest teams do differently. Release it with a public benchmark tool where admins compare their own numbers.

    Original data earns links, press and AI-assistant citations, and it positions OpenWork as the authority on governed AI spend, which is exactly what it sells.

    Risks and how to handle them

    • Not enough opted-in data

      Start with a survey of admins and add gateway data as opt-ins grow.

    Evidence

    • Your websiteEnterprise page: "AI your team wants. Control IT needs." for IT and security leaders: SSO, SCIM, per-team spend limits that warn at 90% and pause at 100%, providers such as AWS Bedrock, and self-hosting; model-gateway support for other coding agents is "still in progress"; every "Book a demo" goes to a calendar page; no named customers, logos or case studies.Source
    • Your websiteRoadmap (updated October 27): shipped since August are AI Gateway spend limits by team, desktop policies, OpenWork Web, automations, audit logs and OpenWork in Slack; planned are a phone app, admin roles, the AI Gateway from other coding agents, approvals mid-run and failure and spend reporting by skill, connection and team; there is no voting, only a "Share feedback" form.Source
    • Your websiteGuides: four pages covering OpenCode, other coding agents via one MCP URL, Ollama (local agents on your own files with no API key) and "How do you keep AI agent costs under control?".Source

Wildcards 1

Unexpected experiments, each with a cheap way to test it.

  • 23

    OpenWork's Product Hunt page has 59 points and no reviews, from a launch about ten months ago when it was a desktop app. Since then it has shipped OpenWork Web, automations, audit logs, spend limits and Slack, and repositioned from a single-vendor alternative to "Your AI workspace. Without vendor lock-in." That is a different product from the one that launched.

    Medium impactMedium effort First result in 1 month Founder & CEO

    First action

    Ask 20 active users to leave a review.

    How to measure it

    Launch-week Cloud orgs

    Now
    –
    Target
    500
    By
    launch + 7 days

    Steps

    1. 1

      Collect reviews

      From active users and contributors.

      Founder & CEO 2 weeks

    2. 2

      Pick the story

      Team sharing and governance.

      Founder & CEO 1 day

    3. 3

      Make assets

      Demo video and screenshots.

      Design Engineer 1 week

    4. 4

      Prepare offer

      Launch-week perk for Team.

      Founder & CEO 1 day

    5. 5

      Launch

      With the community and email list.

      Founder & CEO 1 day

    6. 6

      Follow up

      Onboard every new org within a day.

      Founder & CEO 1 week

    First result
    1 month
    Cost
    Staff time only
    Owner
    Founder & CEO

    Smallest test

    Collect ten reviews on the existing page first.

    Why it works

    Relaunch OpenWork Cloud and the team product as a new launch: a short demo of sharing a template with a team, governance in one screen and the free plan for 5 users. Before launch day, ask the most active users and contributors to leave reviews on the existing page. Use the launch to drive template-gallery and Cloud signups.

    A relaunch with a real story and existing reviews reaches the early-adopter audience a second time, now with something for teams.

    Risks and how to handle them

    • A flat launch

      Launch only after the email list and community are in place.

    Builds on

    Evidence

    • Product Hunt59 points and "No reviews yet", launched in 2026; commenters asked whether it is "safe", what it costs to run compared with Claude Cowork, and whether it works with Slack or Notion.Source
    • Your websiteRoadmap (updated October 27): shipped since August are AI Gateway spend limits by team, desktop policies, OpenWork Web, automations, audit logs and OpenWork in Slack; planned are a phone app, admin roles, the AI Gateway from other coding agents, approvals mid-run and failure and spend reporting by skill, connection and team; there is no voting, only a "Share feedback" form.Source
    • Your websiteSince April 2026 the homepage headline changed from "The open source Claude Cowork for your team" to "Your AI workspace. Without vendor lock-in.", the starting price from $50 a month to $10 per user, and sales booking from Cal.com to a Google appointment schedule; the page title still reads "OpenWork — Free, open-source Claude Cowork alternative".Source

03

How Programmable can help

Where an idea in this playbook is something Programmable builds or runs, it's listed here with what we'd do. Everything else is yours to run.

ProgrammableDigital

Show it, don't just describe it

These ideas need something to show: a demo, a video, an animation, graphics or a deck. Digital turns screenshots or the site into all of them, in the company's own brand.

Make a demo (opens programmable.digital in a new tab)
More example playbooks and reviews

More growth playbooks

Other companies we have researched, each read from what they are hiring for.

  • brightwheel growth strategyAn all-in-one childcare management platform: billing and payments, family communication, attendance and check-in, admissions, a new CRM…22 growth ideas
  • Wave growth strategyMobile money in Africa: free deposits and withdrawals at agents, transfers for 1%, free bill payments and airtime, plus business payments…21 growth ideas
  • Nature Leaf Teas growth strategyA Florida mission-based tea company selling all-natural loose leaf teas, cold brew pouches, local honey, teaware, gift sets and tea…21 growth ideas
  • Weee! growth strategyAmerica's largest online Asian supermarket, with free next-day grocery delivery, a Latino store (MásGusto), a local marketplace for…21 growth ideas
  • LogoAI growth strategyAn AI logo maker and brand studio: logos from a name, living brand kits (colors, type, social assets, videos), mockups, logo animation and…21 growth ideas
  • The Build Inc. growth strategyA boutique brand and growth consultancy for digital-first consumer brands: brand and communications strategy, go-to-market, conversion…20 growth ideas
  • Envoy growth strategyA workplace platform for physical security, safety and compliance: visitor management, threat detection and emergency response (Response)…21 growth ideas
  • Yepoda growth strategyA Berlin-based clean K-Beauty brand making vegan Korean skincare and tinted skincare, sold through its own sites in Europe, the UK and the…22 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on November 10, 2026, and cites them throughout. OpenWork did not commission, review or endorse it, and Programmable is not affiliated with OpenWork. Company and product names are trademarks of their owners, used only to identify them. The ideas are suggestions and opinion, not professional, legal or financial advice, and not statements about how OpenWork runs its business.

Something wrong, or do you represent OpenWork? Email [email protected] and we will correct it or take it down.

30/60/90 roadmap

What to focus on in each month, and the ideas that carry it.

  1. Days 1 to 308 ideas

    Fix the Windows download, add proof, pricing answers, a sales form, an aligned page title, release email, a licence page and brand signals.

  2. Days 31 to 6012 ideas

    Contributor backlog, community, the share-to-Cloud upgrade, the IT policy kit, comparison pages, the switcher campaign, the template gallery, first-run, speed, roadmap follow-up, model launches and a Product Hunt relaunch.

  3. Days 61 to 903 ideas

    Commit to the big bets: a packaged governed pilot, marketplace listings and the AI agent spend report.

The research behind the plan

3 research areas, each summarised in a few lines. Open any of them for the full detail.

Customers

What customers say

Customers value free and open source, with no account needed to start. The most cited problem: vendor lock-in to one model provider.

  • Any model, including local models and their own keys
  • Shadow AI and unmanaged keys inside companies
  • In their words: “no vendor lock-in”

Market

Competitive landscape

3 competitors compared: Claude Cowork, goose and AnythingLLM.

  • Gap to exploit: Governed, any-model agent workspaces that IT can self-host
  • Claude Cowork: One model vendor
  • goose: Community project with no team control plane for sale

Hiring signals

A four-person team is hiring its first two engineers in San Francisco, and the roles show where the product goes next.

  • Founding Engineer: Desktop app, permissions, cloud sync, collaboration and plugin infrastructure; close to 200k downloads. Ideas 1, 2, 9, 11, 15.
  • Design Engineer - SF: First dedicated design hire for a four-person team. Ideas 11, 15, 16.

From the team that wrote this playbook

All products
  • ProgrammableAgencyWorking AI systems in days, not quarters (opens programmable.agency in a new tab)
  • ProgrammableInfoAn agent that finds, researches and asks first (opens programmable.info in a new tab)
  • ProgrammableDigitalScreenshots in, an animated product demo out (opens programmable.digital in a new tab)
  • ProgrammableWatchKnow the day a supplier falls out of compliance (opens programmable.watch in a new tab)
  • ProgrammableTalkGet registrants for your events, and know who’s coming (opens programmable.talk in a new tab)
  • ProgrammablePageText what you need, your customer gets a page that works (opens programmable.page in a new tab)
  • ProgrammableCommunityForums with labelled AI members, never empty (opens programmable.community in a new tab)
ProgrammableMarketing
ExamplesLeaderboardsProductsTermsPrivacyCookiesDPALegalContact

© 2026 AI-Powered Growth Intelligence