ProgrammableMarketing/manufact.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 · 6
  • This quarter · 10
  • Big bets · 3
  • Wildcards · 3
  • How we can help
Generated Nov 10, 2026 1 view
  1. Programmable
  2. Playbooks
  3. Manufact
Manufact logo

growth playbook · Nov 10, 2026

manufact.com (opens in a new tab)

Manufact growth playbook

Diagnosis

Manufact (formerly mcp-use, YC S25, $6.3M seed led by Peak XV) has rare distribution for a five-person team: an SDK with 10.9k GitHub stars, a public leaderboard that has scored more than 3,000 MCP servers, a blog that ships several posts a week, and tool calls it says grew 38% week over week.

Its five open roles show the next steps: an SDR for inbound leads with buying signals, a DevRel who owns partnerships, a forward-deployed engineer for pilots and system integrators, and product and infrastructure engineers.

The gaps are where that reach should turn into accounts: leaderboard pages with no badge or claim option, credits that visitors can't convert into tool calls, templates with no deploy button, no pages comparing Manufact with its closest rivals by name, and customer stories that mostly lack numbers. This playbook puts the leaderboard, pricing clarity and fast follow-up on score runs first, then builds template deploys, partner launches with auth providers and an SI programme, a finance vertical, and data and CI plays as big bets.

Growth Score

7/10

Biggest Opportunity

Turn the 3,000+ public quality reports into the main acquisition loop: every MCP server owner who is scored, shares a badge or runs a fix becomes a signed-up, followed-up lead.

Biggest Risk

Alpic, Smithery (now part of Arcade), Prefect's FastMCP Gateway and general clouds such as Vercel and Cloudflare compete for the same builders, and an MCP-only platform depends on MCP keeping its momentum while some developers argue CLI tools will replace it.

01

Start here

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

  1. 1

    Give each of the 3,000+ scored MCP servers a README badge and a "claim this server" link

    Ship a static badge endpoint for every report and add a "Get your badge" box under the score.

    High impact First result in 2 weeks
  2. 2

    Show what the free credits buy in tool calls, right on the pricing page

    Add one line under each plan showing the tool calls its credit covers.

    High impact First result in 2 weeks
  3. 3

    Make the $5 submission pack the first purchase for teams heading to the ChatGPT and Claude stores

    Add a "Get store-ready" box to the two store-submission blog posts.

    Medium impact First result in 2 weeks

02

Your idea portfolio

22 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.

  • 7 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 · 4Big bets · 8Fill-ins · 7Reconsider · 3
Effort →

Showing 22 of 22 ideas

This week 6

Quick wins with a clear first action.

  • 01

    Manufact's leaderboard has already scored MCP servers on public report pages: 3,066 of the site's 3,524 sitemap URLs are leaderboard pages, and the index lists "1,459 verified MCP servers" plus a Community tab of 3,064. Every report has "Improve", "Share" and "Fix prompt" buttons, but there is no badge to embed, no README snippet, no way for a server's owner to claim the page and no way to follow score changes. The people who own those servers are exactly who Manufact sells to, and today a report is a dead end for them.

    High impactLow effort First result in 2 weeks Developer Advocate & Partners…

    First action

    Ship a static badge endpoint for every report and add a "Get your badge" box under the score.

    How to measure it

    Report-page sign-ups (claims plus Fix with Manufact)

    Now
    –
    Target
    300 claimed servers and 150 sign-ups
    By
    90 days

    Steps

    1. 1

      Build the badge

      A cached SVG per report with score and grade, plus Markdown and HTML snippets.

      Founding Software Engineer (Product / MCP) 2 days

    2. 2

      Add the box

      "Get your badge" and "Claim this server" under the score on every report.

      Founding Software Engineer (Product / MCP) 1 day

    3. 3

      Design verification

      Prove ownership with a well-known file or a token in server metadata; no claims without it.

      Senior Infrastructure Engineer 3 days

    4. 4

      Score-change emails

      Weekly email to claimed owners when a check or grade changes, linking to the fix.

      Founding Software Engineer (Product / MCP) 3 days

    5. 5

      Seed the top grades

      Send A and A+ servers their badge with a short note and a PR-ready snippet.

      Developer Advocate & Partnerships 1 week

    6. 6

      Route claims

      Claims go to the CRM with score gaps; SDR follows up within a day.

      Sales Development Representative 2 days

    7. 7

      Measure

      Badge views, referral clicks, claims, sign-ups and paid conversions.

      Founders (CEO) weekly

    First result
    2 weeks
    Cost
    About two engineer-weeks; no media spend
    Owner
    Developer Advocate & Partnerships

    Smallest test

    Email 50 A-graded server owners a badge snippet and count how many add it within two weeks.

    Why it works

    Add two things to every report page. 1) A badge: a live SVG ("MCP quality 96/100 · A+", scored by Manufact) with copy-paste Markdown for READMEs and docs, linking back to the report. 2) "Claim this server": the owner verifies the server (for example with a well-known file or a tool-metadata token), signs up, and gets a weekly email when their score or a check changes, with a one-click "Fix with Manufact" path to Cloud Inspector, publishing checks and hosting. Send claims to the new SDR as inbound with a buying signal: the score gaps are the opener. Start by emailing the top-graded servers their badge, since A grades are what owners want to show off.

    Badges are how developer tools spread (build status, coverage, package version): each one puts Manufact's name and a link in front of every visitor to that repo, and builds links back to the leaderboard. Claiming turns an anonymous public score into an account with a reason to come back every week, and lower-scored owners get a direct path to the paid checks and hosting that fix the problem.

    Ready to use

    ChecklistLeaderboard badge and claim launch list

    [ ] /leaderboard/<slug>/badge.svg (score, grade, "scored by Manufact"), cached
    [ ] Markdown snippet: [![MCP quality](https://manufact.com/leaderboard/<slug>/badge.svg)](https://manufact.com/leaderboard/<slug>)
    [ ] "Get your badge" box under the score
    [ ] "Claim this server" with ownership proof
    [ ] Weekly score-change email for claimed owners
    [ ] "Fix with Manufact" link to Cloud Inspector and publishing checks
    [ ] Claims into CRM with the three weakest checks
    [ ] Tracking: badge views, referral clicks, claims, sign-ups

    Risks and how to handle them

    • Owners object to a public score or a low grade

      Let claimed owners request a rescore and see exact fixes; show the date of the last review.

    • False claims on popular servers

      Require technical proof of control before a claim is accepted.

    Evidence

    • Your websiteLeaderboard (checked November 9, 2026): title "MCP Quality Checker & Leaderboard", H1 "How good is your MCP?"; "1,459 verified MCP servers, ranked by quality score", a Community tab of 3,064 and a chart of "968 scored MCPs"; a "Score my MCP" form that takes an MCP URL. The page has no badge, embed, README snippet, claim or notify option.Source
    • Your websiteServer report pages (sample checked November 9, 2026): titled "<Server> MCP Server: 96/100 Quality | Manufact", scoring six areas (server metadata, discoverability, protocol, tools, MCP Apps, tool documentation quality) with "Improve", "Share" and "Fix prompt" buttons and a "Score your MCP" link; no badge, embed code, claim option or way for the server's owner to follow changes.Source
    • Your websiteSitemap (3,524 URLs, read November 9, 2026): 3,066 under /leaderboard, 73 blog posts, seven /platform pages, and single pages for /templates, /compare, /startups, /enterprise, /vibe, /builder, /inspector and /stats; no per-competitor comparison pages and no per-template pages.Source
    • HiringSales Development Representative (San Francisco or remote US): handle inbound leads and convert them into qualified demos, follow up warm prospects in the CRM, focus on leads with buying signals "instead of cold blasting", later open new markets and test new acquisition channels.Source
  • 02

    Plans are described in dollars of credit ($5 Free, $10 Pay as you go, $300 Pro) while usage is priced per unit: tool calls at $0.10 per 1K, evals at $0.70 a run, build minutes at $0.07. At launch on Hacker News a developer asked what "credits" means and what pay as you go costs after the limit, and the founder agreed pricing "should be more transparent". $5 is roughly 50,000 tool calls, but nobody is told that.

    High impactLow effort First result in 2 weeks Founders

    First action

    Add one line under each plan showing the tool calls its credit covers.

    How to measure it

    Pricing-to-signup conversion

    Now
    –
    Target
    +20%
    By
    60 days

    Steps

    1. 1

      Agree the conversions

      Credits to tool calls, evals, checks and build minutes.

      Founders (CEO) 1 day

    2. 2

      Add the lines

      One equivalence line per plan.

      Founding Software Engineer (Product / MCP) 1 day

    3. 3

      Build the estimator

      Three inputs, two monthly totals, the break-even.

      Founding Software Engineer (Product / MCP) 3 days

    4. 4

      Sync the docs

      Same numbers on the billing page.

      Developer Advocate & Partnerships 1 day

    5. 5

      Answer in public

      Reply to the launch-thread question with the new page.

      Founders (CEO) 1 hour

    6. 6

      Measure

      Sign-ups and Pro upgrades from the pricing page.

      Founders (CEO) monthly

    First result
    2 weeks
    Cost
    A few engineering days
    Owner
    Founders (CEO)

    Smallest test

    Add the equivalence lines only and compare sign-ups for two weeks.

    Why it works

    Under each plan, translate credits into the units builders think in: "Free: about 50,000 tool calls a month, or 7 evals". Add a small estimator (tool calls a day, evals a week, deploys a week) that shows the monthly cost on Pay as you go and on Pro, and when Pro becomes cheaper. Link the same numbers from the billing docs.

    Developers deploy when they can budget. A visible unit price removes the main question raised at launch and makes the $250 Pro step easier to justify.

    Risks and how to handle them

    • Estimates set wrong expectations

      Label them as examples and link the full rate table.

    Evidence

    • Your websitePricing page: Free $0 with "$5 in monthly credits" and 1 project; Pay as you go ($0 base) with $10 in monthly credits, 10 projects and spending limits; Pro "$250/mo + usage" or $3,000 billed annually, with $300 in monthly credits, unlimited projects, 90-day analytics, cold-start prevention and Slack support; Enterprise custom. Usage rates include tool calls at $0.10 per 1K (first 1M), MCP Evals at $0.70 per run, premium publishing checks at $2.00 per run, a submission pack at $5.00 and build minutes at $0.07 per minute. Allowances are stated as dollars of shared monthly usage.Source
    • Hacker NewsLaunch HN (August 2026, 111 points): a commenter said "Having to sign up to browse available MCPs stopped me dead in my tracks"; another asked what "credits" means and what pay as you go costs after the limit, and the founder agreed pricing "should be more transparent"; others asked which live apps built on the platform are publicly listed and how store discovery works; several argued CLI tools could replace MCP.Source
  • 03

    Manufact sells a submission pack for $5.00 and premium publishing checks for $2.00 a run, and its PredictLeads story is titled "How PredictLeads Avoided Weeks of MCP Marketplace Review". Teams can also connect an existing server by URL to run evals and publish checks without hosting. That is a low-risk first purchase for anyone preparing a store submission, yet it sits in a usage table instead of being sold as a product.

    Medium impactLow effort First result in 2 weeks Founders

    First action

    Add a "Get store-ready" box to the two store-submission blog posts.

    How to measure it

    Paid submission packs

    Now
    –
    Target
    100 a month
    By
    90 days

    Steps

    1. 1

      Define the bundle

      Connect by URL, premium checks, submission pack.

      Founders (CEO) 1 day

    2. 2

      Build a checkout path

      One flow from report or post to purchase.

      Founding Software Engineer (Product / MCP) 3 days

    3. 3

      Add proof

      PredictLeads result on the offer.

      Developer Advocate & Partnerships 1 day

    4. 4

      Place it

      Reports, store-submission posts and pricing.

      Developer Advocate & Partnerships 2 days

    5. 5

      Follow up

      Hosting offer with spend credited back.

      Sales Development Representative ongoing

    6. 6

      Measure

      Pack buyers who go on to host.

      Founders (CEO) monthly

    First result
    2 weeks
    Cost
    Engineering days
    Owner
    Founders (CEO)

    Smallest test

    Add the box to one blog post and count purchases.

    Why it works

    Package "Get store-ready": connect your server URL, run premium publishing checks, get the submission pack, all for under $10, with the PredictLeads result as proof. Put it on the leaderboard reports, the store-submission blog posts and the pricing page. Afterwards, offer hosting with a credit equal to what they spent.

    A small, concrete purchase tied to an urgent deadline (store review) gets a card on file and starts the relationship before the hosting decision.

    Risks and how to handle them

    • Implying store approval

      Say the pack prepares a submission; approval is the store's decision.

    Evidence

    • Your websitePricing page: Free $0 with "$5 in monthly credits" and 1 project; Pay as you go ($0 base) with $10 in monthly credits, 10 projects and spending limits; Pro "$250/mo + usage" or $3,000 billed annually, with $300 in monthly credits, unlimited projects, 90-day analytics, cold-start prevention and Slack support; Enterprise custom. Usage rates include tool calls at $0.10 per 1K (first 1M), MCP Evals at $0.70 per run, premium publishing checks at $2.00 per run, a submission pack at $5.00 and build minutes at $0.07 per minute. Allowances are stated as dollars of shared monthly usage.Source
    • Your websiteCustomer stories: seven (Alpha Vantage, PredictLeads, AgentMail, iAhorro, Meilleurtaux, item, Simera); only three give a number: AgentMail "100,000+ MCP tool calls a week", iAhorro shipped a ChatGPT mortgage simulation app "in 4 days", Meilleurtaux built its simulator "in under two months" as the first French broker on ChatGPT. PredictLeads "Avoided Weeks of MCP Marketplace Review".Source
    • Your websiteDocs: onboarding asks users to Deploy, Connect or Create; "Connect an existing server" lets teams "Paste your MCP URL to run evals and publish checks without hosting"; SDK docs live separately at docs.mcp-use.com.Source
  • 04

    The /compare page sets mcp-use and Manufact against categories (the official SDK, FastMCP, agent frameworks, general clouds, registries, DIY Kubernetes) and has no page for any named MCP hosting rival. The "8 Best MCP Deployment Platforms" post covers Smithery, MintMCP and others but not Alpic, whose homepage pitch ("Build, deploy and distribute ChatGPT apps & MCP servers") is closest to Manufact's.

    High impactLow effort First result in 1 month Developer Advocate & Partners…

    First action

    Write the Alpic page first, since it is the most direct rival.

    How to measure it

    Comparison-page sign-ups

    Now
    –
    Target
    80 a month
    By
    90 days

    Steps

    1. 1

      Check facts

      Each rival's current features and pricing from their own sites.

      Developer Advocate & Partnerships 2 days

    2. 2

      Write pages

      Fair tables, when to choose each, switching help.

      Developer Advocate & Partnerships 1 week

    3. 3

      Add the benchmark

      Link the deployment benchmark and its method.

      Developer Advocate & Partnerships 1 day

    4. 4

      Review

      Claims and trademarks.

      Founders (CEO) 2 days

    5. 5

      Publish and link

      From /compare and the platforms post.

      Founding Software Engineer (Product / MCP) 1 day

    6. 6

      Refresh

      Quarterly, with the date shown.

      Developer Advocate & Partnerships quarterly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Developer Advocate & Partnerships

    Smallest test

    Publish the Alpic page and track sign-ups for a month.

    Why it works

    Write one fair page per closest rival: Alpic, Smithery (now part of Arcade) and Prefect's FastMCP Gateway. Compare hosting, cross-client testing, store submission help, analytics, SDK and pricing, link Manufact's own benchmark, and add a migration guide. Date every page.

    Developers who search a rival's name plus "alternative" are ready to choose. Dated, factual pages also shape how AI assistants describe the category.

    Risks and how to handle them

    • Claims go out of date

      Date each page and cite rivals' own pages.

    Evidence

    • Your websiteCompare page: H1 "mcp-use & Manufact vs Alternatives: August 2026" compares the official MCP SDK, FastMCP, agent frameworks (Vercel AI SDK, LangChain, Mastra), general clouds (Vercel, Render, Fly, Cloudflare Workers), registries and DIY Kubernetes, with no separate page for any named MCP hosting rival.Source
    • Your websiteBlog "The 8 Best MCP Deployment Platforms in 2026" (August 11, 2026) covers Manufact, Cloudflare Workers, Vercel, Smithery, MintMCP, Stainless, Zapier MCP and MCPHosting.io, and links a benchmark scoring Manufact 95, Cloudflare 50, Smithery 44, MintMCP 42 and Vercel 33; Alpic is not among the eight or the platforms listed as left out.Source
    • Competitor researchRivals' own homepages: Alpic, "Build, deploy and distribute ChatGPT apps & MCP servers", with the Skybridge framework, Beacon quality audits and a managed studio; Smithery, "now part of Arcade"; Prefect's FastMCP Gateway, "Free for one developer and up to 200 servers".Source
  • 05

    The SDR role is defined as handling inbound leads with buying signals "instead of cold blasting". Nothing signals intent more clearly than a team pasting its own MCP URL into "Score my MCP": they have a server and want it better. The report already lists the exact failing checks.

    High impactLow effort First result in 1 month Sales Development Representat…

    First action

    Measure today's time from a score run to first contact.

    How to measure it

    Score-to-demo rate

    Now
    –
    Target
    10% of signed-in score runs
    By
    60 days

    Steps

    1. 1

      Measure speed

      Current time to first touch.

      Sales Development Representative 1 day

    2. 2

      Pipe the event

      Score runs into the CRM with the weakest checks.

      Founding Software Engineer (Product / MCP) 2 days

    3. 3

      Write the opener

      Two checks, one fix, one offer.

      Sales Development Representative 1 day

    4. 4

      Tier the response

      Call for company domains, email for others.

      Sales Development Representative 1 day

    5. 5

      Engineer reviews

      15-minute fix sessions.

      Founding Forward Deployed Engineer weekly

    6. 6

      Track

      Demo and paid rates by response time.

      Founders (CEO) weekly

    First result
    1 month
    Cost
    SDR time
    Owner
    Sales Development Representative

    Smallest test

    Hand-reply to one week of score runs within an hour.

    Why it works

    When a signed-in user scores a server, alert the SDR. Within an hour, send a short, specific note: the two weakest checks, the fix prompt, and an offer of a 15-minute review with an engineer. Teams with enterprise domains get a call; others get the self-serve path to Connect, then Deploy.

    Speed and specificity drive developer-tool demo rates, and the report gives a personal opener no cold email can match.

    Risks and how to handle them

    • Developers dislike sales contact

      Lead with the fix, not a meeting request; one follow-up only.

    Evidence

    • HiringSales Development Representative (San Francisco or remote US): handle inbound leads and convert them into qualified demos, follow up warm prospects in the CRM, focus on leads with buying signals "instead of cold blasting", later open new markets and test new acquisition channels.Source
    • Your websiteLeaderboard (checked November 9, 2026): title "MCP Quality Checker & Leaderboard", H1 "How good is your MCP?"; "1,459 verified MCP servers, ranked by quality score", a Community tab of 3,064 and a chart of "968 scored MCPs"; a "Score my MCP" form that takes an MCP URL. The page has no badge, embed, README snippet, claim or notify option.Source
    • Your websiteDocs: onboarding asks users to Deploy, Connect or Create; "Connect an existing server" lets teams "Paste your MCP URL to run evals and publish checks without hosting"; SDK docs live separately at docs.mcp-use.com.Source
  • 06

    The homepage says "10M+" SDK downloads; the YC company description says "over 7M"; the job posts give tool-call growth as both 38% and 35% week over week; the leaderboard shows 1,459 verified servers, 3,064 community entries and "968 scored MCPs" on one page. Enterprise buyers and investors who read closely notice mismatches.

    Medium impactLow effort First result in 1 week Founders

    First action

    List every public number and its source in one sheet.

    How to measure it

    Inconsistent traction figures

    Now
    –
    Target
    0
    By
    2 weeks

    Steps

    1. 1

      Inventory

      Every number on site, YC and posts.

      Founders (CEO) 2 hours

    2. 2

      Pick sources

      npm/PyPI, GitHub, internal tool-call data.

      Founders (CEO) 2 hours

    3. 3

      Update the YC profile and posts

      Same figures and date.

      Founders (CEO) 1 hour

    4. 4

      Update the site

      Homepage, enterprise and leaderboard labels.

      Founding Software Engineer (Product / MCP) half a day

    5. 5

      Link the stats page

      Show live download counts on /stats as the source.

      Founding Software Engineer (Product / MCP) 1 day

    6. 6

      Review monthly

      Refresh with the date.

      Founders (CEO) monthly

    First result
    1 week
    Cost
    $0
    Owner
    Founders (CEO)

    Smallest test

    Fix the YC profile today.

    Why it works

    Pick one dated source of truth for downloads, stars, tool calls and scored servers, update the homepage, YC profile, job posts, enterprise page and leaderboard to match, and label leaderboard counts clearly (verified, community, charted).

    Consistent, dated numbers read as rigour, which matters for a company selling quality scores and SOC 2 to enterprises.

    Risks and how to handle them

    • A smaller number looks worse

      A dated, sourced number is more credible than a larger, unsourced one.

    Evidence

    • Your websiteHomepage: H1 "Manufact: The cloud for production MCP"; "10M+" SDK downloads across Python and TypeScript, "10k+" GitHub stars and "<60s" from git push to a live MCP App or Server; "Used by NASA, LangChain, and thousands of developers"; CTAs "Start for free", "Browse templates" and "Book a call"; sections "Understand how agents use your MCP" and "Catch failures across clients and harnesses".Source
    • HiringFounding Software Engineer (Product / MCP): owns the deploy flow, dashboard, Inspector and docs surface ("You decide what gets built"). The YC posts give both "10M+" and "over 7M" SDK downloads, and tool-call growth as both 38% and 35% week over week.Source
    • Your websiteLeaderboard (checked November 9, 2026): title "MCP Quality Checker & Leaderboard", H1 "How good is your MCP?"; "1,459 verified MCP servers, ranked by quality score", a Community tab of 3,064 and a chart of "968 scored MCPs"; a "Score my MCP" form that takes an MCP URL. The page has no badge, embed, README snippet, claim or notify option.Source

This quarter 10

The core plays for the next three months.

  • 07

    The templates page lists 28 templates, including six auth templates and 13 example projects, but every card links to GitHub or a "Try now" endpoint; there is no deploy button and no page per template. The homepage promises under 60 seconds from git push to live, yet a template visitor has to clone, configure and push before seeing it.

    High impactMedium effort First result in 1 month Founding Software Engineer

    First action

    Add a deploy button to the three starter templates.

    How to measure it

    Deploys started from templates

    Now
    –
    Target
    500 a month
    By
    90 days

    Steps

    1. 1

      Build the flow

      Fork and deploy from a single link.

      Founding Software Engineer (Product / MCP) 1 week

    2. 2

      Starters first

      Starter, MCP Apps and Blank.

      Founding Software Engineer (Product / MCP) 2 days

    3. 3

      Template pages

      One per template with demo and code.

      Developer Advocate & Partnerships 2 weeks

    4. 4

      README buttons

      Same button in every template repo.

      Developer Advocate & Partnerships 2 days

    5. 5

      Onboarding

      Next steps after deploy: test across clients.

      Founding Software Engineer (Product / MCP) 3 days

    6. 6

      Measure

      Deploys and seven-day active projects.

      Founders (CEO) weekly

    First result
    1 month
    Cost
    About three engineer-weeks
    Owner
    Founding Software Engineer (Product / MCP)

    Smallest test

    Add a button to the Starter template and compare deploys with the other cards.

    Why it works

    Give each template a page (/templates/<name>) with what it does, a live demo, the code, and a one-click "Deploy to Manufact" that forks the repo and deploys it. Use the same button in each template's README.

    One-click deploys turn browsing into a running project, the moment a free account becomes active. Template pages also rank for searches like "MCP server Auth0 template".

    Risks and how to handle them

    • Auth templates need secrets

      Prompt for keys in the deploy flow with links to each provider.

    Evidence

    • Your websiteTemplates page: 28 templates (3 starters, 6 auth templates for Convex, Supabase, Better Auth, Keycloak, WorkOS and Auth0, 13 example projects, 6 other frameworks); cards link to GitHub and 22 have a "Try now" live endpoint; no per-template page and no deploy button.Source
    • Your websiteHomepage: H1 "Manufact: The cloud for production MCP"; "10M+" SDK downloads across Python and TypeScript, "10k+" GitHub stars and "<60s" from git push to a live MCP App or Server; "Used by NASA, LangChain, and thousands of developers"; CTAs "Start for free", "Browse templates" and "Book a call"; sections "Understand how agents use your MCP" and "Catch failures across clients and harnesses".Source
    • Your websiteSitemap (3,524 URLs, read November 9, 2026): 3,066 under /leaderboard, 73 blog posts, seven /platform pages, and single pages for /templates, /compare, /startups, /enterprise, /vibe, /builder, /inspector and /stats; no per-competitor comparison pages and no per-template pages.Source
  • 08

    The mcp-use repository has 10.9k stars and 1.5k forks, and its README already has a "Deploy" link and a "Ship to Manufact" section with npm run deploy. The homepage claims 10M+ downloads. The open-source-to-cloud path is the whole business model ("mcp-use as Vercel is to Next.js"), but no public signal shows how many SDK users deploy.

    High impactMedium effort First result in 1 month Founding Software Engineer

    First action

    Add a source parameter to the README deploy link and npm run deploy.

    How to measure it

    Weekly first deploys from SDK paths

    Now
    –
    Target
    +30%
    By
    1 quarter

    Steps

    1. 1

      Tag the paths

      README, CLI, scaffold, docs.

      Founding Software Engineer (Product / MCP) 2 days

    2. 2

      Define the funnel

      Scaffold to paid, with opt-in telemetry.

      Founders (CEO) 1 day

    3. 3

      Build the dashboard

      Weekly cohorts.

      Founding Software Engineer (Product / MCP) 1 week

    4. 4

      Find the drop

      Largest loss between steps.

      Founders (CEO) 1 week

    5. 5

      Fix it

      For example a deploy prompt after scaffolding.

      Founding Software Engineer (Product / MCP) 2 weeks

    6. 6

      Repeat

      Next biggest drop.

      Founding Software Engineer (Product / MCP) monthly

    First result
    1 month
    Cost
    Engineering time
    Owner
    Founding Software Engineer (Product / MCP)

    Smallest test

    Tag the README link and count first deploys for two weeks.

    Why it works

    Tag every path from SDK to Cloud (README link, npm run deploy, create-mcp-use-app, docs) with a source, and build a weekly funnel: scaffolded, first deploy, active after seven days, paid. Fix the step with the largest drop first, for example by prompting a deploy right after scaffolding.

    Open-source companies grow when the jump from local to hosted is short and measured; without numbers the team optimises blind.

    Risks and how to handle them

    • Telemetry upsets open-source users

      Make it opt-in and documented; rely on link tags where possible.

    Evidence

    • GitHubmcp-use repository: 10.9k stars and 1.5k forks; README "The TypeScript framework for MCP" with a "Deploy" link and a "Ship to Manufact" section using npm run deploy; the Releases section is empty.Source
    • Your websiteHomepage: H1 "Manufact: The cloud for production MCP"; "10M+" SDK downloads across Python and TypeScript, "10k+" GitHub stars and "<60s" from git push to a live MCP App or Server; "Used by NASA, LangChain, and thousands of developers"; CTAs "Start for free", "Browse templates" and "Book a call"; sections "Understand how agents use your MCP" and "Catch failures across clients and harnesses".Source
  • 09

    Of seven customer stories, only three carry a result (AgentMail's 100,000+ tool calls a week, iAhorro's four days to a ChatGPT app, Meilleurtaux's under two months). Alpha Vantage, PredictLeads, item and Simera have none, although Manufact's analytics already record tool calls, deploys and checks for each customer.

    Medium impactMedium effort First result in 1 month Developer Advocate & Partners…

    First action

    Pull platform metrics for the four stories without numbers.

    How to measure it

    Stories with numbers

    Now
    –
    Target
    7 of 7
    By
    60 days

    Steps

    1. 1

      Pull metrics

      Tool calls, time to listing, checks caught.

      Founding Software Engineer (Product / MCP) 2 days

    2. 2

      Ask permission

      One email per customer.

      Founders (CEO) 1 week

    3. 3

      Rewrite headlines

      Number first.

      Developer Advocate & Partnerships 3 days

    4. 4

      Update enterprise page

      Show the strongest three.

      Developer Advocate & Partnerships 1 day

    5. 5

      Make it standard

      Every new story needs two numbers.

      Founding Forward Deployed Engineer ongoing

    6. 6

      Measure

      Story page to demo rate.

      Founders (CEO) monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Developer Advocate & Partnerships

    Smallest test

    Add one number to one story and compare demo clicks.

    Why it works

    With each customer's permission, add two numbers per story from platform data: tool calls a week, time from first deploy to store listing, checks caught before review, or clients supported. Put the number in the headline and on the enterprise page.

    Buyers remember numbers, and numbers from the platform itself are more credible than quotes.

    Risks and how to handle them

    • Customers decline to share figures

      Offer ranges or relative figures.

    Evidence

    • Your websiteCustomer stories: seven (Alpha Vantage, PredictLeads, AgentMail, iAhorro, Meilleurtaux, item, Simera); only three give a number: AgentMail "100,000+ MCP tool calls a week", iAhorro shipped a ChatGPT mortgage simulation app "in 4 days", Meilleurtaux built its simulator "in under two months" as the first French broker on ChatGPT. PredictLeads "Avoided Weeks of MCP Marketplace Review".Source
    • Your websiteEnterprise page: H1 "Take MCP from a team experiment to a company capability."; "SOC 2 Type II & ISO 27001 compliance"; Manufact Cloud or bring-your-own-cloud (VPC/private) "scoped with the sales team"; forward-deployed expert support; 15 customer logos including AgentMail, Alpha Vantage, CoinGecko, iAhorro and Meilleurtaux.Source
  • 10

    Manufact maintains OAuth templates for Convex, Supabase, Better Auth, Keycloak, WorkOS and Auth0, and the DevRel hire "owns partnerships with other dev tools / SDKs end-to-end". Auth is where MCP servers most often get stuck, and each provider has its own developer audience looking for a working MCP pattern.

    High impactMedium effort First result in 1 quarter Developer Advocate & Partners…

    First action

    Send the first two providers a one-page co-launch proposal with the template link.

    How to measure it

    Partner-sourced sign-ups

    Now
    –
    Target
    300 a quarter
    By
    2 quarters

    Steps

    1. 1

      Rank partners

      Audience overlap and template usage.

      Developer Advocate & Partnerships 2 days

    2. 2

      Write the proposal

      Guide, template, livestream, listing.

      Developer Advocate & Partnerships 2 days

    3. 3

      Pitch

      First two providers.

      Developer Advocate & Partnerships 2 weeks

    4. 4

      Build the guide

      Co-written, tested end to end.

      Developer Advocate & Partnerships 2 weeks

    5. 5

      Launch

      Both blogs, both newsletters, a livestream.

      Developer Advocate & Partnerships 1 day

    6. 6

      Measure

      Tagged sign-ups and deploys.

      Founders (CEO) monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Developer Advocate & Partnerships

    Smallest test

    Publish one auth guide and ask the provider to share it.

    Why it works

    Offer each provider a joint package: a co-written guide ("Add <provider> auth to your MCP server"), the template with a Deploy to Manufact button, a joint livestream, and a listing in the provider's integrations or docs. Start with the two whose developer audiences overlap most with MCP builders.

    Partners lend their reach for free when the content solves their users' problem, and the template already exists.

    Risks and how to handle them

    • Partners move slowly

      Publish Manufact's side first and invite them to co-sign.

    Builds on

    Evidence

    • Your websiteTemplates page: 28 templates (3 starters, 6 auth templates for Convex, Supabase, Better Auth, Keycloak, WorkOS and Auth0, 13 example projects, 6 other frameworks); cards link to GitHub and 22 have a "Try now" live endpoint; no per-template page and no deploy button.Source
    • HiringDeveloper Advocate & Partnerships: "owns partnerships with other dev tools / SDKs end-to-end", builds demos and integrations with partner products, makes videos, GitHub examples, blog posts and docs, and represents Manufact at conferences and meetups; goal: "A big portion of global MCP tool calls will run on Manufact"; tool calls on Manufact grew 38% week over week since June; $6.3M seed led by Peak XV.Source
  • 11

    The forward-deployed engineer post includes "partner engineering" with system integrators and consultants deploying MCP apps for their clients. Alpic sells a managed studio that delivers production apps "in days or weeks". Manufact has no partner page, so agencies that want to build on it have no listing, pricing or certification.

    High impactMedium effort First result in 1 quarter Founding Forward Deployed Eng…

    First action

    List agencies already deploying on Manufact from project data.

    How to measure it

    Partner-built client projects

    Now
    –
    Target
    20 live
    By
    2 quarters

    Steps

    1. 1

      Find early partners

      Agencies already on the platform.

      Founding Forward Deployed Engineer 1 week

    2. 2

      Set terms

      Margin, referral and support levels.

      Founders (CEO) 1 week

    3. 3

      Certification

      Based on publishing checks and a sample app.

      Founding Forward Deployed Engineer 2 weeks

    4. 4

      Partner page

      Directory and apply form.

      Developer Advocate & Partnerships 1 week

    5. 5

      Route projects

      Small inbound projects to partners.

      Sales Development Representative ongoing

    6. 6

      Review

      Quality and client retention.

      Founding Forward Deployed Engineer quarterly

    First result
    1 quarter
    Cost
    Staff time and partner margin
    Owner
    Founding Forward Deployed Engineer

    Smallest test

    Sign three agencies and refer them one project each.

    Why it works

    Launch a small programme: a partner page, a directory of certified agencies, a partner workspace with client projects and margin on Pro, a short certification based on the publishing checks, and referral of inbound projects too small for Manufact's own team.

    Agencies multiply delivery capacity for a five-person team and bring enterprise clients who need someone to build the app, without Manufact running a services business.

    Risks and how to handle them

    • Poor partner work hurts the brand

      Certify and require projects to pass publishing checks.

    Evidence

    • HiringFounding Forward Deployed Engineer: technical discovery and pilots with success criteria, implementation to production, post-launch adoption, "partner engineering" with developer tools, system integrators and consultants, and turning repeated work into templates, docs and product improvements.Source
    • Competitor researchRivals' own homepages: Alpic, "Build, deploy and distribute ChatGPT apps & MCP servers", with the Skybridge framework, Beacon quality audits and a managed studio; Smithery, "now part of Arcade"; Prefect's FastMCP Gateway, "Free for one developer and up to 200 servers".Source
    • Your websiteSitemap (3,524 URLs, read November 9, 2026): 3,066 under /leaderboard, 73 blog posts, seven /platform pages, and single pages for /templates, /compare, /startups, /enterprise, /vibe, /builder, /inspector and /stats; no per-competitor comparison pages and no per-template pages.Source
  • 12

    Manufact's strongest stories are in finance: iAhorro shipped a mortgage simulation app on ChatGPT in four days, Meilleurtaux became the first French broker on ChatGPT, and Alpha Vantage and CoinGecko are customers. That is a vertical with clear use cases (simulators, rates, market data), compliance needs Manufact already covers (SOC 2 Type II, ISO 27001), and competitors who will copy the first movers.

    High impactMedium effort First result in 1 quarter Founders

    First action

    Ask iAhorro and Meilleurtaux to join a joint webinar or quote.

    How to measure it

    Demos from financial services

    Now
    –
    Target
    15 a quarter
    By
    2 quarters

    Steps

    1. 1

      Collect proof

      Timelines and results from the finance stories.

      Founding Forward Deployed Engineer 1 week

    2. 2

      Write the page

      Use cases, proof, compliance, launch plan.

      Developer Advocate & Partnerships 1 week

    3. 3

      Build a demo app

      A sample rate simulator template.

      Founding Forward Deployed Engineer 2 weeks

    4. 4

      Outbound list

      Similar brokers and data providers, described by segment.

      Sales Development Representative 1 week

    5. 5

      Webinar

      With one customer.

      Founders (CEO) 1 month

    6. 6

      Measure

      Demos and pilots from finance.

      Founders (CEO) monthly

    First result
    1 quarter
    Cost
    Staff time only
    Owner
    Founders (CEO)

    Smallest test

    Publish the page and send it to 20 lenders and brokers.

    Why it works

    Build one industry page: what a financial ChatGPT app does, the three stories, the compliance and data-residency answers, and a launch plan with timelines from the real projects. Use it for outbound to similar brokers and data providers, and run a webinar with one of the customers.

    Regulated buyers want proof from peers; a page that bundles finance stories and compliance answers shortens the evaluation.

    Risks and how to handle them

    • Regulatory questions slow deals

      Prepare a security and data FAQ from the Trust Center.

    Builds on

    Evidence

    • Your websiteCustomer stories: seven (Alpha Vantage, PredictLeads, AgentMail, iAhorro, Meilleurtaux, item, Simera); only three give a number: AgentMail "100,000+ MCP tool calls a week", iAhorro shipped a ChatGPT mortgage simulation app "in 4 days", Meilleurtaux built its simulator "in under two months" as the first French broker on ChatGPT. PredictLeads "Avoided Weeks of MCP Marketplace Review".Source
    • Your websiteEnterprise page: H1 "Take MCP from a team experiment to a company capability."; "SOC 2 Type II & ISO 27001 compliance"; Manufact Cloud or bring-your-own-cloud (VPC/private) "scoped with the sales team"; forward-deployed expert support; 15 customer logos including AgentMail, Alpha Vantage, CoinGecko, iAhorro and Meilleurtaux.Source
  • 13

    "Understand how agents use your MCP" is a headline promise, and plans include 30 to 365 days of analytics, but the value sits in a dashboard developers visit only when something breaks. A weekly digest brings them back.

    Medium impactMedium effort First result in 1 month Founding Software Engineer

    First action

    Send a hand-made digest to the ten most active projects.

    How to measure it

    Projects active week over week

    Now
    –
    Target
    +15%
    By
    1 quarter

    Steps

    1. 1

      Pick the metrics

      Calls by client, failing tools, errors.

      Founding Software Engineer (Product / MCP) 1 day

    2. 2

      Hand-made pilot

      Ten projects.

      Founding Forward Deployed Engineer 1 week

    3. 3

      Automate

      Monday email job.

      Founding Software Engineer (Product / MCP) 1 week

    4. 4

      Add a fix

      One suggestion from checks.

      Founding Software Engineer (Product / MCP) 3 days

    5. 5

      Add usage line

      Credit used and Pro break-even.

      Founding Software Engineer (Product / MCP) 1 day

    6. 6

      Measure

      Opens, return visits, upgrades.

      Founders (CEO) monthly

    First result
    1 month
    Cost
    Engineering time
    Owner
    Founding Software Engineer (Product / MCP)

    Smallest test

    Hand-send ten digests and compare return visits.

    Why it works

    Email each project owner every Monday: tool calls by client, top and failing tools, new errors, and one suggested fix from the checks. Include a usage-versus-credit line so teams see when Pro makes sense.

    Regular, useful reports keep the product in view, catch problems before users churn, and set up upgrades with real numbers.

    Risks and how to handle them

    • Email fatigue

      Skip weeks with no change; one-click opt-out.

    Evidence

    • Your websiteHomepage: H1 "Manufact: The cloud for production MCP"; "10M+" SDK downloads across Python and TypeScript, "10k+" GitHub stars and "<60s" from git push to a live MCP App or Server; "Used by NASA, LangChain, and thousands of developers"; CTAs "Start for free", "Browse templates" and "Book a call"; sections "Understand how agents use your MCP" and "Catch failures across clients and harnesses".Source
    • Your websitePricing page: Free $0 with "$5 in monthly credits" and 1 project; Pay as you go ($0 base) with $10 in monthly credits, 10 projects and spending limits; Pro "$250/mo + usage" or $3,000 billed annually, with $300 in monthly credits, unlimited projects, 90-day analytics, cold-start prevention and Slack support; Enterprise custom. Usage rates include tool calls at $0.10 per 1K (first 1M), MCP Evals at $0.70 per run, premium publishing checks at $2.00 per run, a submission pack at $5.00 and build minutes at $0.07 per minute. Allowances are stated as dollars of shared monthly usage.Source
  • 14

    The blog has 73 posts, 17 of them dated between October 29 and November 6, 2026 (cookbooks, store-submission guides, monthly updates), but there is no newsletter signup and no RSS feed. Readers who arrive from search leave with no way to hear about the next post.

    Medium impactLow effort First result in 1 month Developer Advocate & Partners…

    First action

    Publish an RSS feed of the blog.

    How to measure it

    Subscribers

    Now
    –
    Target
    3,000
    By
    2 quarters

    Steps

    1. 1

      Add RSS

      Feed for /blog.

      Founding Software Engineer (Product / MCP) half a day

    2. 2

      Pick a tool

      Email platform with double opt-in.

      Developer Advocate & Partnerships 1 day

    3. 3

      Add signups

      Posts, leaderboard, docs.

      Founding Software Engineer (Product / MCP) 1 day

    4. 4

      Write issue one

      Cookbooks and store news.

      Developer Advocate & Partnerships 2 days

    5. 5

      Send fortnightly

      Fixed day.

      Developer Advocate & Partnerships ongoing

    6. 6

      Measure

      Subscribers, clicks, sign-ups.

      Founders (CEO) monthly

    First result
    1 month
    Cost
    Under $100 a month
    Owner
    Developer Advocate & Partnerships

    Smallest test

    Add a signup box to the five most-read posts.

    Why it works

    Add an RSS feed and a fortnightly email: new cookbooks, store-policy changes, the month's product updates and a leaderboard highlight. Put the signup on every post, the leaderboard and the docs.

    Owned email turns one-off search visits into a repeat audience, and store-policy news gives developers a reason to subscribe.

    Risks and how to handle them

    • Not enough time to write

      Assemble from posts already published.

    Evidence

    • Your websiteBlog: 73 posts, 17 of them dated October 29 to November 6, 2026; no newsletter signup and no RSS feed found.Source
    • Your websiteSitemap (3,524 URLs, read November 9, 2026): 3,066 under /leaderboard, 73 blog posts, seven /platform pages, and single pages for /templates, /compare, /startups, /enterprise, /vibe, /builder, /inspector and /stats; no per-competitor comparison pages and no per-template pages.Source
  • 15

    The startups page is for teams with less than $2M in funding but states no credit amount ("Your offer will specify its included usage"), so founders can't judge if applying is worth it. Manufact is itself a YC company with investors who back early teams.

    Medium impactLow effort First result in 1 month Founders

    First action

    Decide the credit amount and put it in the page headline.

    How to measure it

    Startup applications

    Now
    –
    Target
    100 a quarter
    By
    2 quarters

    Steps

    1. 1

      Set the offer

      Credits, length, eligibility.

      Founders (CEO) 1 day

    2. 2

      Update the page

      Number in the headline.

      Founding Software Engineer (Product / MCP) half a day

    3. 3

      Batch deals

      YC and two other accelerators.

      Founders (CEO) 1 month

    4. 4

      Onboarding

      Startup-specific first-week email.

      Developer Advocate & Partnerships 2 days

    5. 5

      Expiry follow-up

      Upgrade path before credits end.

      Sales Development Representative ongoing

    6. 6

      Measure

      Applications and paid conversions.

      Founders (CEO) quarterly

    First result
    1 month
    Cost
    Credits given away
    Owner
    Founders (CEO)

    Smallest test

    Post the number on the page and compare applications.

    Why it works

    Publish a clear offer (for example, a set amount of credits or months of Pro) with eligibility, and arrange batch-wide deals with YC and other accelerators, starting with those where Manufact's investors are active. Track which startups grow into paying Pro accounts.

    A stated number gets more applications than a vague one, and accelerator batches are dense groups of teams building AI products right now.

    Risks and how to handle them

    • Credit cost with no conversion

      Cap credits and require a live deploy to unlock the rest.

    Evidence

    • Your websiteStartups page: for startups "with less than $2M in total funding", bootstrapped teams welcome, reviewed individually; the page states no credit amount ("Your offer will specify its included usage and billing terms").Source
    • HiringFive open roles on Y Combinator (team of 5, YC S25, San Francisco): Founding Forward Deployed Engineer, Founding Software Engineer (Product / MCP), Sales Development Representative, Senior Infrastructure Engineer, and Developer Advocate & Partnerships.Source
  • 16

    The /vibe page says "Vibe your ChatGPT Apps", is marked beta, and its only call to action is "Go to Dashboard", with no description of what it does. Code Builder generates Python agent code with mcp-use but sits apart from the deploy path. Both are in the sitemap and can be found by search, yet a visitor can't tell what they get.

    Medium impactLow effort First result in 2 weeks Founding Software Engineer

    First action

    Write the one-sentence description and the example.

    How to measure it

    Sign-ups from Vibe and Builder

    Now
    –
    Target
    3x current
    By
    60 days

    Steps

    1. 1

      Define the promise

      What Vibe builds, for whom.

      Founders (CEO) 1 day

    2. 2

      Write the page

      Example, steps, CTA.

      Developer Advocate & Partnerships 2 days

    3. 3

      Signed-out path

      "Start free" instead of dashboard.

      Founding Software Engineer (Product / MCP) 1 day

    4. 4

      Builder next step

      Deploy link after code is generated.

      Founding Software Engineer (Product / MCP) 1 day

    5. 5

      Link in

      From templates and homepage.

      Founding Software Engineer (Product / MCP) half a day

    6. 6

      Measure

      Sign-ups by page.

      Founders (CEO) monthly

    First result
    2 weeks
    Cost
    A few days
    Owner
    Founding Software Engineer (Product / MCP)

    Smallest test

    Add one explanatory paragraph and a sign-up button to /vibe.

    Why it works

    Give Vibe a real landing page: one sentence on what it builds, a short example (prompt in, ChatGPT app out), who it is for, and "Start free" for signed-out visitors. On Code Builder, add a next step: deploy the generated agent's servers on Manufact.

    Visitors only sign in to something they understand; a clear page turns a dead end into an entry point for less technical builders.

    Risks and how to handle them

    • Beta not ready for more users

      Use a waitlist until capacity is there.

    Evidence

    • Your websiteVibe and Builder pages: "Vibe your ChatGPT Apps" is a beta page whose only call to action is "Go to Dashboard", with no description of what it does; Code Builder generates Python agent code using mcp-use.Source
    • Your websiteSitemap (3,524 URLs, read November 9, 2026): 3,066 under /leaderboard, 73 blog posts, seven /platform pages, and single pages for /templates, /compare, /startups, /enterprise, /vibe, /builder, /inspector and /stats; no per-competitor comparison pages and no per-template pages.Source

Big bets 3

Larger plays that can change the trajectory.

  • 17

    Manufact scores public MCP servers across six areas (metadata, discoverability, protocol, tools, MCP Apps, tool documentation) and already publishes a deployment benchmark. No one else holds a dataset of this size on MCP server quality. Aggregate findings (how many servers declare output schemas, which checks fail most, how quality differs by category) would be cited by press, analysts and the client platforms.

    High impactHigh effort First result in 1 quarter Founders

    First action

    Export the current scores and find the five most surprising patterns.

    How to measure it

    Report email captures

    Now
    –
    Target
    2,000
    By
    60 days after launch

    Steps

    1. 1

      Export data

      Scores and check results by category.

      Founding Software Engineer (Product / MCP) 1 week

    2. 2

      Write method

      How scoring works, limits.

      Founders (CEO) 1 week

    3. 3

      Find the story

      Five headline findings.

      Founders (CEO) 1 week

    4. 4

      Design the report

      Charts and a public summary.

      Developer Advocate & Partnerships 3 weeks

    5. 5

      Press and talks

      Embargoed briefings and a conference slot.

      Founders (CEO) 1 month

    6. 6

      Launch

      Gated PDF, open summary, newsletter.

      Developer Advocate & Partnerships 1 day

    7. 7

      Update quarterly

      Trend lines.

      Founding Software Engineer (Product / MCP) quarterly

    First result
    1 quarter
    Cost
    Staff time plus design
    Owner
    Founders (CEO)

    Smallest test

    Publish one blog post with three aggregate findings.

    Why it works

    Publish an annual report with quarterly updates: aggregate scores by category, the most common failures and how to fix them, trends over time and a methodology section. Gate the full PDF behind an email, launch with a talk at an MCP conference and give journalists early access.

    Original data earns links and press that blog posts can't, positions Manufact as the quality standard, and feeds the badge loop.

    Risks and how to handle them

    • Naming low-scoring servers causes friction

      Report aggregates only; never single out weak servers.

    Builds on

    Evidence

    • Your websiteServer report pages (sample checked November 9, 2026): titled "<Server> MCP Server: 96/100 Quality | Manufact", scoring six areas (server metadata, discoverability, protocol, tools, MCP Apps, tool documentation quality) with "Improve", "Share" and "Fix prompt" buttons and a "Score your MCP" link; no badge, embed code, claim option or way for the server's owner to follow changes.Source
    • Your websiteLeaderboard (checked November 9, 2026): title "MCP Quality Checker & Leaderboard", H1 "How good is your MCP?"; "1,459 verified MCP servers, ranked by quality score", a Community tab of 3,064 and a chart of "968 scored MCPs"; a "Score my MCP" form that takes an MCP URL. The page has no badge, embed, README snippet, claim or notify option.Source
    • Your websiteBlog "The 8 Best MCP Deployment Platforms in 2026" (August 11, 2026) covers Manufact, Cloudflare Workers, Vercel, Smithery, MintMCP, Stainless, Zapier MCP and MCPHosting.io, and links a benchmark scoring Manufact 95, Cloudflare 50, Smithery 44, MintMCP 42 and Vercel 33; Alpic is not among the eight or the platforms listed as left out.Source
  • 18

    Manufact already runs publishing checks ($0.10 a run on the publishing checklist), evals and cross-client tests, and teams can run checks on servers they don't host. Today those checks run when someone visits the dashboard or leaderboard. Running them on every pull request would put Manufact into the daily workflow of every MCP repo, including those hosted elsewhere.

    High impactHigh effort First result in 1 quarter Founding Software Engineer

    First action

    Prototype the Action against the mcp-use example servers.

    How to measure it

    Active repos with the check

    Now
    –
    Target
    1,000
    By
    2 quarters

    Steps

    1. 1

      Prototype

      Run checks on a PR and comment.

      Founding Software Engineer (Product / MCP) 2 weeks

    2. 2

      Ephemeral servers

      Spin up the PR build for checks.

      Senior Infrastructure Engineer 3 weeks

    3. 3

      Pricing

      Free public cap, metered private.

      Founders (CEO) 1 week

    4. 4

      Marketplace listing

      Publish the Action with docs.

      Developer Advocate & Partnerships 1 week

    5. 5

      Default in scaffold

      Include in create-mcp-use-app.

      Founding Software Engineer (Product / MCP) 2 days

    6. 6

      Launch

      Blog, demo video, newsletter.

      Developer Advocate & Partnerships 1 week

    7. 7

      Measure

      Repos, runs and conversions to hosting.

      Founders (CEO) monthly

    First result
    1 quarter
    Cost
    About six engineer-weeks plus compute for free runs
    Owner
    Founding Software Engineer (Product / MCP)

    Smallest test

    Run checks on pull requests in Manufact's own example repos for a month.

    Why it works

    Ship a GitHub Action that runs publishing checks (and optionally evals) on each pull request, posts the score change as a comment and fails the build on regressions. Free for public repos with a monthly cap; private repos use the metered rates. Add it to the mcp-use scaffold by default.

    CI tools spread repo by repo and are hard to remove once teams rely on them. A visible score in every pull request markets Manufact to every contributor and leads naturally to hosting.

    Risks and how to handle them

    • Compute cost of free runs

      Cap free runs and cache unchanged results.

    Evidence

    • Your websitePricing page: Free $0 with "$5 in monthly credits" and 1 project; Pay as you go ($0 base) with $10 in monthly credits, 10 projects and spending limits; Pro "$250/mo + usage" or $3,000 billed annually, with $300 in monthly credits, unlimited projects, 90-day analytics, cold-start prevention and Slack support; Enterprise custom. Usage rates include tool calls at $0.10 per 1K (first 1M), MCP Evals at $0.70 per run, premium publishing checks at $2.00 per run, a submission pack at $5.00 and build minutes at $0.07 per minute. Allowances are stated as dollars of shared monthly usage.Source
    • Your websiteDocs: onboarding asks users to Deploy, Connect or Create; "Connect an existing server" lets teams "Paste your MCP URL to run evals and publish checks without hosting"; SDK docs live separately at docs.mcp-use.com.Source
    • GitHubmcp-use repository: 10.9k stars and 1.5k forks; README "The TypeScript framework for MCP" with a "Deploy" link and a "Ship to Manufact" section using npm run deploy; the Releases section is empty.Source
  • 19

    The enterprise page offers Manufact Cloud or a bring-your-own-cloud (VPC/private) deployment "scoped with the sales team", and Enterprise lists self-hosting. Every scoping takes founder and engineer time, and buyers in finance and other regulated sectors can't see what they would get. The Senior Infrastructure Engineer Manufact is hiring will build "tooling and blueprints" so customers can deploy Manufact into their own private clouds on AWS, Azure or GCP: the job post describes the product, but the website doesn't yet.

    High impactHigh effort First result in 2 quarters Senior Infrastructure Engineer

    First action

    Write the reference architecture for one cloud.

    How to measure it

    Qualified BYOC opportunities

    Now
    –
    Target
    5
    By
    2 quarters

    Steps

    1. 1

      Pick the first cloud

      Where current prospects run.

      Senior Infrastructure Engineer 1 week

    2. 2

      Reference architecture

      Responsibilities and network design.

      Senior Infrastructure Engineer 2 weeks

    3. 3

      Installer

      Repeatable deploy and upgrades.

      Senior Infrastructure Engineer 6 weeks

    4. 4

      Security pack

      From the Trust Center.

      Founders (CEO) 2 weeks

    5. 5

      Design partners

      Two deployments.

      Founding Forward Deployed Engineer 1 quarter

    6. 6

      Publish the offer

      Page and starting price.

      Founders (CEO) 1 week

    First result
    2 quarters
    Cost
    Two engineer-quarters
    Owner
    Senior Infrastructure Engineer

    Smallest test

    Show the reference architecture to three enterprise prospects.

    Why it works

    Turn BYOC into a product: supported clouds and regions, a reference architecture, what Manufact manages versus the customer, upgrade cadence, a security pack from the Trust Center, and a starting price. Ship an installer for one cloud first and run two design-partner deployments.

    A defined offer shortens enterprise cycles and lets the SDR qualify regulated buyers without a founder call.

    Risks and how to handle them

    • Support burden across many environments

      Support one cloud first and a fixed set of versions.

    Evidence

    • Your websiteEnterprise page: H1 "Take MCP from a team experiment to a company capability."; "SOC 2 Type II & ISO 27001 compliance"; Manufact Cloud or bring-your-own-cloud (VPC/private) "scoped with the sales team"; forward-deployed expert support; 15 customer logos including AgentMail, Alpha Vantage, CoinGecko, iAhorro and Meilleurtaux.Source
    • HiringFive open roles on Y Combinator (team of 5, YC S25, San Francisco): Founding Forward Deployed Engineer, Founding Software Engineer (Product / MCP), Sales Development Representative, Senior Infrastructure Engineer, and Developer Advocate & Partnerships.Source
    • HiringSenior Infrastructure Engineer: build and scale the cloud that hosts customer MCP servers with monitoring and real-time alerting, own the release pipeline, and create "tooling and blueprints" so customers can deploy Manufact into their own private clouds on AWS, Azure or GCP with isolated data boundaries.Source

Wildcards 3

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

  • 20

    At launch on Hacker News a commenter asked which live apps built on the platform are publicly listed beyond Manufact's own, and another asked how store discovery works. The customer stories show apps in ChatGPT (iAhorro, Meilleurtaux), but there is no gallery where a visitor can try them.

    Medium impactLow effort First result in 1 month Developer Advocate & Partners…

    First action

    Ask the ten largest projects for permission to list them.

    How to measure it

    Apps in the showcase

    Now
    –
    Target
    40
    By
    1 quarter

    Steps

    1. 1

      Ask builders

      Opt-in from active projects.

      Developer Advocate & Partnerships 1 week

    2. 2

      Build the page

      Cards, filters by client and category.

      Founding Software Engineer (Product / MCP) 1 week

    3. 3

      Badge

      "Built with Manufact" for builders' sites.

      Founding Software Engineer (Product / MCP) 2 days

    4. 4

      Feature

      One app per newsletter.

      Developer Advocate & Partnerships ongoing

    5. 5

      Submission form

      Self-serve listing.

      Founding Software Engineer (Product / MCP) 2 days

    6. 6

      Measure

      Referral traffic and sign-ups.

      Founders (CEO) monthly

    First result
    1 month
    Cost
    Staff time only
    Owner
    Developer Advocate & Partnerships

    Smallest test

    Publish a page with the five apps from public customer stories.

    Why it works

    Build /showcase: live ChatGPT apps and Claude connectors running on Manufact, each with a screenshot, the client stores it is in, a link to try it and the builder's site. Offer builders a "Built with Manufact" badge and a feature in the newsletter in return for listing.

    Working apps are the best proof that the platform delivers, and builders share pages that feature them.

    Risks and how to handle them

    • Few builders opt in

      Start with customer-story companies who are already public.

    Evidence

    • Hacker NewsLaunch HN (August 2026, 111 points): a commenter said "Having to sign up to browse available MCPs stopped me dead in my tracks"; another asked what "credits" means and what pay as you go costs after the limit, and the founder agreed pricing "should be more transparent"; others asked which live apps built on the platform are publicly listed and how store discovery works; several argued CLI tools could replace MCP.Source
    • Your websiteCustomer stories: seven (Alpha Vantage, PredictLeads, AgentMail, iAhorro, Meilleurtaux, item, Simera); only three give a number: AgentMail "100,000+ MCP tool calls a week", iAhorro shipped a ChatGPT mortgage simulation app "in 4 days", Meilleurtaux built its simulator "in under two months" as the first French broker on ChatGPT. PredictLeads "Avoided Weeks of MCP Marketplace Review".Source
  • 21

    Manufact ran or joined 11 events between March and October 2026, from MCP Apps Night and an MCP V2 release party in San Francisco to AGNTCon + MCPCon in San Jose and Amsterdam. The events page now says "Nothing on the calendar right now." The DevRel role includes representing Manufact at meetups.

    Medium impactMedium effort First result in 1 month Developer Advocate & Partners…

    First action

    Book the next three dates and put them on the events page.

    How to measure it

    Sign-ups from events

    Now
    –
    Target
    50 a month
    By
    1 quarter

    Steps

    1. 1

      Set the format

      Three demos, one talk, drinks.

      Developer Advocate & Partnerships 1 day

    2. 2

      Book dates

      Three months ahead.

      Developer Advocate & Partnerships 1 week

    3. 3

      Co-host

      One partner per event.

      Developer Advocate & Partnerships ongoing

    4. 4

      Record

      Each demo as a short video.

      Developer Advocate & Partnerships monthly

    5. 5

      Follow up

      Attendee list to CRM with consent.

      Sales Development Representative monthly

    6. 6

      Measure

      Attendance, sign-ups, views.

      Founders (CEO) monthly

    First result
    1 month
    Cost
    Venue and food, often shared with partners
    Owner
    Developer Advocate & Partnerships

    Smallest test

    Run one event with a partner and count sign-ups.

    Why it works

    Own one recurring format: a monthly MCP Apps Night with three live demos of apps built that month, recorded and cut into short videos for the blog and newsletter. Co-host with a partner each time (auth providers, dev tools) to share the audience.

    A regular event becomes the local meeting point for MCP builders, produces content every month, and gives partners an easy way to co-market.

    Risks and how to handle them

    • Low turnout

      Co-host with partners who bring their own lists.

    Evidence

    • Your websiteEvents page: 11 past events from March to October 2026 (San Francisco, San Jose, New York, Amsterdam, Turin), including MCP Apps Night and AGNTCon + MCPCon; "Nothing on the calendar right now."Source
    • HiringDeveloper Advocate & Partnerships: "owns partnerships with other dev tools / SDKs end-to-end", builds demos and integrations with partner products, makes videos, GitHub examples, blog posts and docs, and represents Manufact at conferences and meetups; goal: "A big portion of global MCP tool calls will run on Manufact"; tool calls on Manufact grew 38% week over week since June; $6.3M seed led by Peak XV.Source
  • 22

    The loudest objection in the launch thread was not about Manufact but about MCP itself: "MCP is a deadend. CLI use is the future", "MCP is a protocol, that's all", and worries about context bloat. Manufact's whole market depends on that argument, and it sees more MCP traffic than almost anyone, with tool calls it says grew 38% week over week.

    Medium impactLow effort First result in 2 weeks Founders

    First action

    Collect the objections from the launch thread and sales calls.

    How to measure it

    Referring domains to the essay

    Now
    –
    Target
    30
    By
    60 days

    Steps

    1. 1

      Collect objections

      Launch thread and sales calls.

      Founders (CEO) 1 day

    2. 2

      Pull data

      Aggregate usage patterns, no customer names.

      Founding Software Engineer (Product / MCP) 3 days

    3. 3

      Write it

      When each approach wins.

      Founders (CEO) 1 week

    4. 4

      Publish and share

      Blog, Hacker News, X, newsletter.

      Developer Advocate & Partnerships 1 day

    5. 5

      Link it

      Compare page and sales follow-ups.

      Sales Development Representative 1 day

    6. 6

      Talk version

      At the next meetup or conference.

      Founders (CEO) 1 month

    First result
    2 weeks
    Cost
    Staff time only
    Owner
    Founders (CEO)

    Smallest test

    Post a short thread with three data points and gauge the response.

    Why it works

    Publish a clear, fair essay and talk: when a CLI or skill is the better choice, when MCP wins (auth, distribution through ChatGPT and Claude stores, non-developer users), and what Manufact's aggregate data shows about real usage. Link it from the compare page and use it in sales.

    Taking on the category's main doubt openly earns trust with sceptical developers and gives the team the reference answer to share.

    Risks and how to handle them

    • Looks self-serving

      Say plainly where a CLI is the better tool.

    Evidence

    • Hacker NewsLaunch HN (August 2026, 111 points): a commenter said "Having to sign up to browse available MCPs stopped me dead in my tracks"; another asked what "credits" means and what pay as you go costs after the limit, and the founder agreed pricing "should be more transparent"; others asked which live apps built on the platform are publicly listed and how store discovery works; several argued CLI tools could replace MCP.Source
    • HiringDeveloper Advocate & Partnerships: "owns partnerships with other dev tools / SDKs end-to-end", builds demos and integrations with partner products, makes videos, GitHub examples, blog posts and docs, and represents Manufact at conferences and meetups; goal: "A big portion of global MCP tool calls will run on Manufact"; tool calls on Manufact grew 38% week over week since June; $6.3M seed led by Peak XV.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.

  • IrisAgent growth strategyAn AI customer support platform that runs beside existing helpdesks (Zendesk, Salesforce, Intercom, Freshdesk): AI agents for chat, email…21 growth ideas
  • Huscarl growth strategyAn AI-native actuarial advisory for corporate self-insurance: captive feasibility studies, pricing and underwriting for hard-to-price…22 growth ideas
  • Deepgram growth strategyVoice AI platform: real-time and batch speech-to-text, text-to-speech and a unified Voice Agent API (Flux and Nova models), sold as cloud…21 growth ideas
  • SARELLY growth strategyA Mexican makeup and fashion-accessories brand co-founded by creator Anna Sarelly: waterproof mascaras, lips, palettes and kits, plus bags…22 growth ideas
  • Mintive growth strategyInfrastructure for putting AI agents to work safely: an MCP gateway with role-based bundles, an LLM gateway, Mint Guard runtime…21 growth ideas
  • Mews growth strategyA cloud hospitality operating system: property management, point of sale, revenue management, embedded payments, guest technology and a…21 growth ideas
  • Horizon3 growth strategyNodeZero, an autonomous pentesting platform that safely runs real attacks in production across internal, external, cloud, Kubernetes…23 growth ideas
  • ClearGov growth strategyCloud software for public sector finance: strategic planning, capital, personnel and operating budgeting, digital budget books, financial…20 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. Manufact did not commission, review or endorse it, and Programmable is not affiliated with Manufact. 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 Manufact runs its business.

Something wrong, or do you represent Manufact? 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 307 ideas

    Leaderboard badges and claims, credits shown as tool calls, the submission pack as a first purchase, named comparison pages, one-hour follow-up on score runs, consistent numbers and a clear Vibe page.

  2. Days 31 to 609 ideas

    Template deploys, SDK-to-Cloud tracking, numbers in every story, auth-provider co-launches, an SI programme, a finance vertical, weekly agent digests, a newsletter and a concrete startup offer.

  3. Days 61 to 906 ideas

    Commit to the big bets (State of MCP report, publishing checks in CI, packaged bring-your-own-cloud) and run the wildcards: a live-app showcase, a monthly meetup and a clear answer to "MCP or CLI?".

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 going from git push to a live MCP server in under a minute. The most cited problem: unclear store submission rules for ChatGPT and Claude.

  • Catching store-review problems before submission
  • Servers that work in one client and fail in another
  • In their words: “MCP Apps”

Market

Competitive landscape

4 competitors compared: Alpic, Smithery (part of Arcade), FastMCP Gateway (Prefect) and Cloudflare Workers and Vercel.

  • Gap to exploit: A trusted public quality standard for MCP servers
  • Alpic: Smaller open-source footprint than mcp-use
  • Smithery (part of Arcade): Agent-side focus rather than server builders

Hiring signals

A five-person YC S25 team is hiring for the step from open-source reach to revenue.

  • Sales Development Representative (SDR): Inbound leads with buying signals, not cold outreach. Ideas 1, 5.
  • Developer Advocate & Partnerships (DevRel): Owns dev-tool partnerships end to end; content and events. Ideas 4, 10, 21.
  • Founding Forward Deployed Engineer: Pilots, adoption and partner engineering with system integrators. Ideas 11, 12.

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 firstNew accounts get 100,000 free enrichment credits with code BETA100K (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)
  • ProgrammableMediaAn AI note taker, interpreter and webinar host for your calls (opens programmable.media 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