Orthogonal growth playbook

Assign every uncategorized API to one primary and up to two secondary categories, add the few categories the catalog now needs (for example Social Data, Verification, Voice and Phone, Financial Data, Jobs Data, Compute), and make a category required in provider onboarding so the count never drifts back. Then check that the MCP and REST search return the newly tagged APIs for category queries, and that each category page links to all of them.

Most buyers, human or agent, start from a task (find an email, search the web), not a brand name. Today the best-known APIs for several tasks are invisible from the task pages, so the catalog undersells itself and pushes visitors to rivals that look bigger. Tagging is a one-week job with no engineering risk, and it improves every later idea that relies on categories (category landing pages, comparisons, selection).

[ ] Serper -> Search (and Web Data)
[ ] Icypeas -> Email Finder (and People Data)
[ ] Leadsforge -> People Data (and Email Finder)
[ ] TinyFish -> Browser Automation (and Web Data)
[ ] Olostep -> Web Data (and Scrape)
[ ] Scrape Creators -> new Social Data category
[ ] Abstract -> Identity Verification
[ ] Company Enrich -> Company Data
[ ] Remaining 54 APIs tagged
[ ] Category required in onboarding
[ ] MCP and REST category search checked
[ ] Uncategorized count at 0

• Too many small categories

  Cap at about 20 and use secondary tags for the long tail.

• Providers disagree with their category

  Let them suggest a change during review.

• Your websiteCatalog (checked November 10, 2026): "87 APIs · 1,774 endpoints"; categories AI Search (9), Brand Data (2), Browser Automation (2), Company Data (6), Email Finder (4), Identity Verification (1), People Data (6), Video (1), Weather (1), Web Data (8), Scrape (3), Search (2), LLM (2), Image (2) and Uncategorized (62); the only sort is "Most called"; cards show "Verified publisher" and "Reliable" badges and a from-price, but no call volume, latency, uptime, ratings or reviews.Source
• Your website/catalog/uncategorized ("page 1 of 8", checked twice on November 10, 2026) lists Olostep, Icypeas, Serper (a Google Search API for web, image, video, news, maps, shopping and scholar results), Leadsforge, TinyFish (AI-driven browser automation and web search), Company Enrich, Scrape Creators and Abstract. Olostep, Icypeas and Serper are among the first cards on the "Most called" catalog view with no category tag, while the Search category shows 2 APIs, Email Finder 4 and Browser Automation 2.Source
• Your websiteHomepage (checked November 10, 2026): H1 "Get your agent the data it needs."; "Access data and services through one connection. Pay for the calls you make, without a separate subscription to each provider."; "87 APIs" and "1,774 endpoints"; "Explore all 14 categories"; per-call prices from "$0.002" to "from $0.50"; CTAs "Explore APIs", "Connect an agent", "Get your API key" and "Book a call"; no testimonials, customer logos or customer results on the page.Source
• Your websiteBlog and use cases: 59 posts, the recent ones mostly partner spotlights (TinyFish on September 23, 2026, then Extruct, Akta, Financial Datasets, FullEnrich) with TypeScript, x402 and MPP code and a "Try It Today" free-credit call to action; Breaknine is the only customer case study among recent posts. The docs list five use cases, all go-to-market: lead research, competitor brand and pricing research, people search, account-based research and news monitoring.Source

Publish the guide at that URL (or redirect it to the current guide), add it to the sitemap, and add a link check to the build that fails when llms.txt, skill.md or the docs link to a missing page.

Supply is half of a marketplace, and the agent-led listing path is a differentiator only if it works end to end. A redirect is minutes of work.

• Other agent-facing links break later

  Keep the build-time link check.

• Your websitellms.txt (checked November 10, 2026): indexes 59 per-API skill files against 87 APIs in the catalog (TinyFish, featured in a September 23 partner post, is not in it); it points providers to "List your API: https://www.orthogonal.com/docs/list-your-api", which returned 404 Not Found on two checks that day; it mentions a one-time $1 starter credit and that "Top-up packages carry a 5% processing fee".Source
• Your websiteSell page: "Reach developers and AI agents with your API."; listing is "Free" with no monthly minimum; providers "set your provider price" and marketplace pricing terms are shared in the onboarding workspace after approval; payouts for the previous month are "initiated on the 15th (UTC)"; five-step onboarding with the Orth onboarding agent; a "Get verified" link with no stated requirements; no buyer or call statistics.Source
• Your websiteSitemap (263 URLs, checked November 10, 2026): 85 /publisher pages, 85 /discover/<slug>/skill.md files, 59 blog posts and 18 docs pages; no /catalog/<category> pages; no /compare, /vs or /customers pages and no top-level use-case pages (only /docs/use-cases); the site has no changelog and no newsletter signup.Source

Build llms.txt and skill.md from the same catalog data the site uses, at deploy time, with each API's categories and from-price, and list the machine-readable /discover and /openapi.json endpoints at the top. Add a test that the count in llms.txt equals the catalog count.

For an agent-first product, llms.txt is the storefront. Keeping it generated means every new provider is discoverable by agents on day one, which is also what providers are paying attention to.

• File grows too long for agents

  Keep one line per API and link to per-API skill files.

• Your websitellms.txt (checked November 10, 2026): indexes 59 per-API skill files against 87 APIs in the catalog (TinyFish, featured in a September 23 partner post, is not in it); it points providers to "List your API: https://www.orthogonal.com/docs/list-your-api", which returned 404 Not Found on two checks that day; it mentions a one-time $1 starter credit and that "Top-up packages carry a 5% processing fee".Source
• Your websiteSitemap (263 URLs, checked November 10, 2026): 85 /publisher pages, 85 /discover/<slug>/skill.md files, 59 blog posts and 18 docs pages; no /catalog/<category> pages; no /compare, /vs or /customers pages and no top-level use-case pages (only /docs/use-cases); the site has no changelog and no newsletter signup.Source
• Your websiteCatalog (checked November 10, 2026): "87 APIs · 1,774 endpoints"; categories AI Search (9), Brand Data (2), Browser Automation (2), Company Data (6), Email Finder (4), Identity Verification (1), People Data (6), Video (1), Weather (1), Web Data (8), Scrape (3), Search (2), LLM (2), Image (2) and Uncategorized (62); the only sort is "Most called"; cards show "Verified publisher" and "Reliable" badges and a from-price, but no call volume, latency, uptime, ratings or reviews.Source

Pull the live API and endpoint counts into the homepage, about page, sell page and llms.txt from the catalog, update the YC profile to the current number, and add a short "catalog growth" line (35+ in June, 87 in November) as a proof point. Date-stamp counts in older posts rather than editing history.

Growth is the story investors, providers and buyers want to hear, and a consistent, rising number tells it. It also stops assistants describing Orthogonal as smaller than it is.

• Counts drop when APIs are delisted

  Show endpoints and APIs together; both are honest.

• Your websiteCatalog (checked November 10, 2026): "87 APIs · 1,774 endpoints"; categories AI Search (9), Brand Data (2), Browser Automation (2), Company Data (6), Email Finder (4), Identity Verification (1), People Data (6), Video (1), Weather (1), Web Data (8), Scrape (3), Search (2), LLM (2), Image (2) and Uncategorized (62); the only sort is "Most called"; cards show "Verified publisher" and "Reliable" badges and a from-price, but no call volume, latency, uptime, ratings or reviews.Source
• Your websiteSeed announcement (June 25, 2026): $4.3 million led by Pantera Capital, with Y Combinator, Pioneer Fund, Decasonic, Blast, Outbound and Surreal by Premise; "already live in production" with GTM teams and AI-native startups; more than 35 APIs at the time, including Apollo, People Data Labs and Coresignal; fiat, x402 and MPP payment rails; the money funds "expansion of Orthogonal's engineering and go-to-market teams".Source
• Hacker NewsShow HN (September 1, 2026, 12 points, no comments): "one integration to discover, access, and pay for APIs" with 50+ providers and 700+ endpoints; it started with go-to-market APIs and is expanding; "The problem we're working on now is API selection", choosing providers by price, latency, reliability and result quality; the founders ask whether people would trust an agent with a budget to choose its own API provider.Source
• Y CombinatorYC profile: Winter 2026, team size 2, "APIs weren't built for agents. Orthogonal fixes that.", and agents can discover, access and pay for hundreds of APIs; Orthogonal's own about page says it "connects dozens of verified providers".Source

Add every category page to the sitemap, give each a task-led title and description, a short intro, the price range across providers, a one-line MCP example and links to the providers in it. Fix label and URL mismatches (Company Data lives at /catalog/company-search, Web Data at /catalog/web-search) so titles match what people search.

Category pages answer the comparison searches that single-provider pages and partner posts cannot, and a marketplace is the only kind of site that can honestly show several rivals on one page.

• Thin pages

  Only index categories with three or more APIs after idea 1.

• #1 Put the 62 "Uncategorized" APIs into categories, so the catalog's 14 categories show the whole catalog

• Your websiteSitemap (263 URLs, checked November 10, 2026): 85 /publisher pages, 85 /discover/<slug>/skill.md files, 59 blog posts and 18 docs pages; no /catalog/<category> pages; no /compare, /vs or /customers pages and no top-level use-case pages (only /docs/use-cases); the site has no changelog and no newsletter signup.Source
• Your websiteCatalog (checked November 10, 2026): "87 APIs · 1,774 endpoints"; categories AI Search (9), Brand Data (2), Browser Automation (2), Company Data (6), Email Finder (4), Identity Verification (1), People Data (6), Video (1), Weather (1), Web Data (8), Scrape (3), Search (2), LLM (2), Image (2) and Uncategorized (62); the only sort is "Most called"; cards show "Verified publisher" and "Reliable" badges and a from-price, but no call volume, latency, uptime, ratings or reviews.Source
• HiringGrowth Lead (San Francisco, in person, 5+ years): help with "accelerating revenue growth beyond the '0-to-1 phase'", build GTM strategy with the founders "to hit our revenue targets", run multi-channel demand generation, design A/B tests to improve conversion rates, research audiences and competitors, and measure performance.Source

Add a proof strip under the hero with the Breaknine numbers and a link to the case study, then collect two or three more short results (time saved per integration, providers swapped without code changes) from production customers who agree to be named.

Developers trust specific numbers from a similar team more than claims. One result also answers the unspoken question for a seed-stage marketplace: does anyone use this in production?

• Grounding metric is unfamiliar

  Explain it in plain words: evidence from a source other than the company itself.

• Your websiteHomepage (checked November 10, 2026): H1 "Get your agent the data it needs."; "Access data and services through one connection. Pay for the calls you make, without a separate subscription to each provider."; "87 APIs" and "1,774 endpoints"; "Explore all 14 categories"; per-call prices from "$0.002" to "from $0.50"; CTAs "Explore APIs", "Connect an agent", "Get your API key" and "Book a call"; no testimonials, customer logos or customer results on the page.Source
• Your websiteBreaknine case study (August 4, 2026): a tool that turns venture investors' CRM data into outreach moved its news discovery onto Orthogonal, using Fiber AI, Apollo, Serper, Bytemine, Company Enrich, PredictLeads, Fundable and Aviato; in its first two weeks it surfaced 622 updates from 233 source domains, and 539 of 622 (about 87%) were independently grounded, against "close to zero" before.Source
• Your websiteSeed announcement (June 25, 2026): $4.3 million led by Pantera Capital, with Y Combinator, Pioneer Fund, Decasonic, Blast, Outbound and Surreal by Premise; "already live in production" with GTM teams and AI-native startups; more than 35 APIs at the time, including Apollo, People Data Labs and Coresignal; fiat, x402 and MPP payment rails; the money funds "expansion of Orthogonal's engineering and go-to-market teams".Source

On signup, in the dashboard and in the agent signup response, show a "what $1 gets you" menu of five tasks with exact costs and a copy-paste prompt for each, chosen from cheap, reliable endpoints. Track which first task leads to a top-up.

A concrete first win inside the free credit is the fastest path to the first top-up, and it steers new users away from expensive endpoints that would drain the credit in one call.

• Users only use cheap endpoints

  Show one premium task as the next step after the first top-up.

• Your websitePricing docs: "You pay only for what you use. No subscriptions."; illustrative tiers Basic $0.001 to $0.01, Standard $0.01 to $0.10, Premium $0.10 to $1.00; each response returns "priceCents"; organisation API keys bill the organisation; an agent's own account "gets a one-time $1 starter credit (one per phone number)"; wallet payments over x402, MPP or Nanopayments. The page states no platform fee and does not say whether a failed call is charged.Source
• Your websiteApollo listing: "Verified", "9 endpoints" priced $0.0020 to $0.0500 per call, a "skill.md (for agents)" link, SDK and CLI quickstart, "More in Company Search" related APIs and a "Get your key" button; no call counts, latency, uptime, reviews or comparison with buying from the provider directly. The sitemap lists 85 publisher pages like it.Source
• Your websiteHomepage (checked November 10, 2026): H1 "Get your agent the data it needs."; "Access data and services through one connection. Pay for the calls you make, without a separate subscription to each provider."; "87 APIs" and "1,774 endpoints"; "Explore all 14 categories"; per-call prices from "$0.002" to "from $0.50"; CTAs "Explore APIs", "Connect an agent", "Get your API key" and "Book a call"; no testimonials, customer logos or customer results on the page.Source

Show 30-day calls (or a usage band), median and p95 latency, and success rate on each provider page and in a side-by-side table on each category page, and return the same fields in the MCP and REST search so agents can choose. Let providers see their own figures first, then publish.

Real performance data turns the catalog into a decision tool, gives good providers a reason to promote their listing, and is an asset rivals cannot copy without the call volume.

• Providers object to low scores

  Preview first and publish bands rather than raw numbers for small samples.

• #1 Put the 62 "Uncategorized" APIs into categories, so the catalog's 14 categories show the whole catalog

• Hacker NewsShow HN (September 1, 2026, 12 points, no comments): "one integration to discover, access, and pay for APIs" with 50+ providers and 700+ endpoints; it started with go-to-market APIs and is expanding; "The problem we're working on now is API selection", choosing providers by price, latency, reliability and result quality; the founders ask whether people would trust an agent with a budget to choose its own API provider.Source
• Your websiteCatalog (checked November 10, 2026): "87 APIs · 1,774 endpoints"; categories AI Search (9), Brand Data (2), Browser Automation (2), Company Data (6), Email Finder (4), Identity Verification (1), People Data (6), Video (1), Weather (1), Web Data (8), Scrape (3), Search (2), LLM (2), Image (2) and Uncategorized (62); the only sort is "Most called"; cards show "Verified publisher" and "Reliable" badges and a from-price, but no call volume, latency, uptime, ratings or reviews.Source
• Your websiteApollo listing: "Verified", "9 endpoints" priced $0.0020 to $0.0500 per call, a "skill.md (for agents)" link, SDK and CLI quickstart, "More in Company Search" related APIs and a "Get your key" button; no call counts, latency, uptime, reviews or comparison with buying from the provider directly. The sitemap lists 85 publisher pages like it.Source
• Competitor researchRivals' homepages: Composio, "Give your AI secure access to 1,700+ apps in minutes", with a free Hobby plan of 100,000 tool calls a month and some tools billed per call; Smithery, "Connect agents to thousands of tools and services. Auth, credentials, and sessions handled for you."; OpenRouter, the model gateway Orthogonal's founders compare it to, publishes public rankings of usage by tokens processed.See the research Source

Each month, publish the top APIs by calls overall and per category, the fastest movers and the newest entries, with anonymised totals or bands. Send it to providers (who will share their ranking), post it on X and LinkedIn, and keep an archive page.

A recurring, data-led page earns links and mentions that partner spotlights do not, gives providers a reason to promote Orthogonal, and turns usage data into authority.

• Reveals a customer's usage

  Use bands and minimum sample sizes.

• Your websiteCatalog (checked November 10, 2026): "87 APIs · 1,774 endpoints"; categories AI Search (9), Brand Data (2), Browser Automation (2), Company Data (6), Email Finder (4), Identity Verification (1), People Data (6), Video (1), Weather (1), Web Data (8), Scrape (3), Search (2), LLM (2), Image (2) and Uncategorized (62); the only sort is "Most called"; cards show "Verified publisher" and "Reliable" badges and a from-price, but no call volume, latency, uptime, ratings or reviews.Source
• Competitor researchRivals' homepages: Composio, "Give your AI secure access to 1,700+ apps in minutes", with a free Hobby plan of 100,000 tool calls a month and some tools billed per call; Smithery, "Connect agents to thousands of tools and services. Auth, credentials, and sessions handled for you."; OpenRouter, the model gateway Orthogonal's founders compare it to, publishes public rankings of usage by tokens processed.See the research Source
• Your websiteBlog and use cases: 59 posts, the recent ones mostly partner spotlights (TinyFish on September 23, 2026, then Extruct, Akta, Financial Datasets, FullEnrich) with TypeScript, x402 and MPP code and a "Try It Today" free-credit call to action; Breaknine is the only customer case study among recent posts. The docs list five use cases, all go-to-market: lead research, competitor brand and pricing research, people search, account-based research and news monitoring.Source

Publish recipes such as "Monitor news on a list of companies", "Find and verify a work email", "Build an account brief" and "Scrape a site that has no API": the APIs used, a copy-paste MCP prompt, the code, and the measured cost per run (for example per 100 companies). Link each recipe from the matching category page and provider pages.

Task pages match search intent and show the unique value of a multi-provider catalog: combining sources in one flow with one bill. The cost per run removes the main question before a top-up.

• Costs change with provider prices

  Recompute cost per run automatically each month.

• Your websiteBlog and use cases: 59 posts, the recent ones mostly partner spotlights (TinyFish on September 23, 2026, then Extruct, Akta, Financial Datasets, FullEnrich) with TypeScript, x402 and MPP code and a "Try It Today" free-credit call to action; Breaknine is the only customer case study among recent posts. The docs list five use cases, all go-to-market: lead research, competitor brand and pricing research, people search, account-based research and news monitoring.Source
• Your websiteBreaknine case study (August 4, 2026): a tool that turns venture investors' CRM data into outreach moved its news discovery onto Orthogonal, using Fiber AI, Apollo, Serper, Bytemine, Company Enrich, PredictLeads, Fundable and Aviato; in its first two weeks it surfaced 622 updates from 233 source domains, and 539 of 622 (about 87%) were independently grounded, against "close to zero" before.Source
• Your websiteSitemap (263 URLs, checked November 10, 2026): 85 /publisher pages, 85 /discover/<slug>/skill.md files, 59 blog posts and 18 docs pages; no /catalog/<category> pages; no /compare, /vs or /customers pages and no top-level use-case pages (only /docs/use-cases); the site has no changelog and no newsletter signup.Source

Ship a kit for each provider: a badge and a deep link to their Orthogonal page, a snippet for their docs ("Call us from any MCP client with one key"), a co-branded recipe, and a monthly stats email. Track signups by provider referral link and feature the providers who send the most traffic.

In a two-sided marketplace the supply side is also a distribution channel. A link from a provider's docs reaches developers who already want that data and makes Orthogonal the easy way to try it.

• Providers fear channel conflict

  Show that Orthogonal brings incremental, per-call buyers they would not sign directly.

• Your websiteSell page: "Reach developers and AI agents with your API."; listing is "Free" with no monthly minimum; providers "set your provider price" and marketplace pricing terms are shared in the onboarding workspace after approval; payouts for the previous month are "initiated on the 15th (UTC)"; five-step onboarding with the Orth onboarding agent; a "Get verified" link with no stated requirements; no buyer or call statistics.Source
• Your websiteSitemap (263 URLs, checked November 10, 2026): 85 /publisher pages, 85 /discover/<slug>/skill.md files, 59 blog posts and 18 docs pages; no /catalog/<category> pages; no /compare, /vs or /customers pages and no top-level use-case pages (only /docs/use-cases); the site has no changelog and no newsletter signup.Source
• Your websiteBlog and use cases: 59 posts, the recent ones mostly partner spotlights (TinyFish on September 23, 2026, then Extruct, Akta, Financial Datasets, FullEnrich) with TypeScript, x402 and MPP code and a "Try It Today" free-credit call to action; Breaknine is the only customer case study among recent posts. The docs list five use cases, all go-to-market: lead research, competitor brand and pricing research, people search, account-based research and news monitoring.Source

State the marketplace take (or the range), how provider price maps to buyer price, payout method and threshold, and the verification checklist with what the badge gives (placement, the "Verified publisher" label). Add a short FAQ and an earnings example.

Clear terms raise application rates and shorten the founders' sales calls with providers; verification becomes a goal providers work toward.

• Locks in terms too early

  Publish as current terms with a review date.

• Your websiteSell page: "Reach developers and AI agents with your API."; listing is "Free" with no monthly minimum; providers "set your provider price" and marketplace pricing terms are shared in the onboarding workspace after approval; payouts for the previous month are "initiated on the 15th (UTC)"; five-step onboarding with the Orth onboarding agent; a "Get verified" link with no stated requirements; no buyer or call statistics.Source
• Your websiteCatalog (checked November 10, 2026): "87 APIs · 1,774 endpoints"; categories AI Search (9), Brand Data (2), Browser Automation (2), Company Data (6), Email Finder (4), Identity Verification (1), People Data (6), Video (1), Weather (1), Web Data (8), Scrape (3), Search (2), LLM (2), Image (2) and Uncategorized (62); the only sort is "Most called"; cards show "Verified publisher" and "Reliable" badges and a from-price, but no call volume, latency, uptime, ratings or reviews.Source

Send email and webhook alerts at 50%, 80% and 100% of each cap and at a low-balance threshold, add an optional auto top-up, and return a "remaining budget" field with each call so agents can plan. Show recent near-misses in the dashboard.

Usage-based products lose customers when production breaks unexpectedly. Warnings and auto top-up keep agents running and turn a refusal into a top-up.

• Alert fatigue

  Let users pick thresholds and channels.

• Your websiteSpend limits docs: a daily cap per account or organisation and one per API key, counted since midnight UTC; per-key allowed-API lists ("A call to an API outside either list is refused with HTTP 403"); refusal codes DAILY_ACCOUNT_LIMIT_EXCEEDED and DAILY_KEY_LIMIT_EXCEEDED; no alerts before a cap or a low balance is reached are described.Source
• Your websitePricing docs: "You pay only for what you use. No subscriptions."; illustrative tiers Basic $0.001 to $0.01, Standard $0.01 to $0.10, Premium $0.10 to $1.00; each response returns "priceCents"; organisation API keys bill the organisation; an agent's own account "gets a one-time $1 starter credit (one per phone number)"; wallet payments over x402, MPP or Nanopayments. The page states no platform fee and does not say whether a failed call is charged.Source

Write a one-paragraph policy (for example: provider and network errors are never charged; empty results follow the provider's rule, shown on each endpoint) and show a "charged on empty result: yes/no" flag per endpoint. Report failed-call totals in the usage view.

Trust is the main barrier to giving an agent a wallet. A clear rule removes a common objection in sales calls and makes per-call pricing feel safe.

• Providers bill empty results

  Show the provider's rule rather than absorbing the cost.

• Your websitePricing docs: "You pay only for what you use. No subscriptions."; illustrative tiers Basic $0.001 to $0.01, Standard $0.01 to $0.10, Premium $0.10 to $1.00; each response returns "priceCents"; organisation API keys bill the organisation; an agent's own account "gets a one-time $1 starter credit (one per phone number)"; wallet payments over x402, MPP or Nanopayments. The page states no platform fee and does not say whether a failed call is charged.Source
• Hacker NewsShow HN (September 1, 2026, 12 points, no comments): "one integration to discover, access, and pay for APIs" with 50+ providers and 700+ endpoints; it started with go-to-market APIs and is expanding; "The problem we're working on now is API selection", choosing providers by price, latency, reliability and result quality; the founders ask whether people would trust an agent with a budget to choose its own API provider.Source

Build a calculator: pick tasks and monthly volumes, see the per-call cost through Orthogonal from live catalog prices, and a field for the team's current subscriptions. Offer the result as a shareable link and a prompt for a "Book a call" for larger volumes.

It turns the core message into a number a buyer can take to their manager, and it qualifies larger teams for the founders' calls.

• Unfair comparison with providers' plans

  Let users enter their own subscription costs instead of quoting rivals' prices.

• Your websiteHomepage (checked November 10, 2026): H1 "Get your agent the data it needs."; "Access data and services through one connection. Pay for the calls you make, without a separate subscription to each provider."; "87 APIs" and "1,774 endpoints"; "Explore all 14 categories"; per-call prices from "$0.002" to "from $0.50"; CTAs "Explore APIs", "Connect an agent", "Get your API key" and "Book a call"; no testimonials, customer logos or customer results on the page.Source
• Your websiteSeed announcement (June 25, 2026): $4.3 million led by Pantera Capital, with Y Combinator, Pioneer Fund, Decasonic, Blast, Outbound and Surreal by Premise; "already live in production" with GTM teams and AI-native startups; more than 35 APIs at the time, including Apollo, People Data Labs and Coresignal; fiat, x402 and MPP payment rails; the money funds "expansion of Orthogonal's engineering and go-to-market teams".Source
• Your websiteApollo listing: "Verified", "9 endpoints" priced $0.0020 to $0.0500 per call, a "skill.md (for agents)" link, SDK and CLI quickstart, "More in Company Search" related APIs and a "Get your key" button; no call counts, latency, uptime, reviews or comparison with buying from the provider directly. The sitemap lists 85 publisher pages like it.Source

Publish a changelog page generated from catalog changes (new APIs, new endpoints, price drops) and send a short weekly email to users with a one-line prompt to try each new API. Let agents read the same feed.

For a catalog, new supply is the main reason to come back. A weekly note re-activates dormant keys at almost no cost.

• Thin weeks

  Send every two weeks when fewer than three changes.

• Your websiteSitemap (263 URLs, checked November 10, 2026): 85 /publisher pages, 85 /discover/<slug>/skill.md files, 59 blog posts and 18 docs pages; no /catalog/<category> pages; no /compare, /vs or /customers pages and no top-level use-case pages (only /docs/use-cases); the site has no changelog and no newsletter signup.Source
• Your websiteSeed announcement (June 25, 2026): $4.3 million led by Pantera Capital, with Y Combinator, Pioneer Fund, Decasonic, Blast, Outbound and Surreal by Premise; "already live in production" with GTM teams and AI-native startups; more than 35 APIs at the time, including Apollo, People Data Labs and Coresignal; fiat, x402 and MPP payment rails; the money funds "expansion of Orthogonal's engineering and go-to-market teams".Source
• Your websiteCatalog (checked November 10, 2026): "87 APIs · 1,774 endpoints"; categories AI Search (9), Brand Data (2), Browser Automation (2), Company Data (6), Email Finder (4), Identity Verification (1), People Data (6), Video (1), Weather (1), Web Data (8), Scrape (3), Search (2), LLM (2), Image (2) and Uncategorized (62); the only sort is "Most called"; cards show "Verified publisher" and "Reliable" badges and a from-price, but no call volume, latency, uptime, ratings or reviews.Source

Create a Teams page: shared balance, per-member keys with their own caps and allowed APIs, monthly invoicing for larger balances (which can also avoid card processing fees), usage by member, and a named contact. Price it on prepaid commitment rather than a seat fee, to keep the no-subscription promise.

Team buyers spend more and churn less than individuals, and the controls they need already exist. Packaging turns features into a reason to commit a larger prepaid balance.

• Looks like a subscription

  Keep it prepaid usage with no seat fees.

• #13 Warn before an agent hits its daily cap or runs out of balance

• Your websitePricing docs: "You pay only for what you use. No subscriptions."; illustrative tiers Basic $0.001 to $0.01, Standard $0.01 to $0.10, Premium $0.10 to $1.00; each response returns "priceCents"; organisation API keys bill the organisation; an agent's own account "gets a one-time $1 starter credit (one per phone number)"; wallet payments over x402, MPP or Nanopayments. The page states no platform fee and does not say whether a failed call is charged.Source
• Your websiteSpend limits docs: a daily cap per account or organisation and one per API key, counted since midnight UTC; per-key allowed-API lists ("A call to an API outside either list is refused with HTTP 403"); refusal codes DAILY_ACCOUNT_LIMIT_EXCEEDED and DAILY_KEY_LIMIT_EXCEEDED; no alerts before a cap or a low balance is reached are described.Source
• Your websitellms.txt (checked November 10, 2026): indexes 59 per-API skill files against 87 APIs in the catalog (TinyFish, featured in a September 23 partner post, is not in it); it points providers to "List your API: https://www.orthogonal.com/docs/list-your-api", which returned 404 Not Found on two checks that day; it mentions a one-time $1 starter credit and that "Top-up packages carry a 5% processing fee".Source
• Your websiteSeed announcement (June 25, 2026): $4.3 million led by Pantera Capital, with Y Combinator, Pioneer Fund, Decasonic, Blast, Outbound and Surreal by Premise; "already live in production" with GTM teams and AI-native startups; more than 35 APIs at the time, including Apollo, People Data Labs and Coresignal; fiat, x402 and MPP payment rails; the money funds "expansion of Orthogonal's engineering and go-to-market teams".Source

Pick one adjacent vertical where agents already need paid data, for example financial and market data (Financial Datasets is already listed) or web automation (TinyFish, Olostep, scraping APIs), and make it complete: five to eight providers, recipes, a category landing page and a launch with those providers. Measure whether it brings a new kind of buyer.

Depth in one new vertical beats one API in many. It widens the market for the next round and tests whether the model works outside GTM.

• Distracts from the GTM core

  One vertical at a time, with a go/no-go review.

• #1 Put the 62 "Uncategorized" APIs into categories, so the catalog's 14 categories show the whole catalog

• Hacker NewsShow HN (September 1, 2026, 12 points, no comments): "one integration to discover, access, and pay for APIs" with 50+ providers and 700+ endpoints; it started with go-to-market APIs and is expanding; "The problem we're working on now is API selection", choosing providers by price, latency, reliability and result quality; the founders ask whether people would trust an agent with a budget to choose its own API provider.Source
• Your websiteCatalog (checked November 10, 2026): "87 APIs · 1,774 endpoints"; categories AI Search (9), Brand Data (2), Browser Automation (2), Company Data (6), Email Finder (4), Identity Verification (1), People Data (6), Video (1), Weather (1), Web Data (8), Scrape (3), Search (2), LLM (2), Image (2) and Uncategorized (62); the only sort is "Most called"; cards show "Verified publisher" and "Reliable" badges and a from-price, but no call volume, latency, uptime, ratings or reviews.Source
• Your websiteBlog and use cases: 59 posts, the recent ones mostly partner spotlights (TinyFish on September 23, 2026, then Extruct, Akta, Financial Datasets, FullEnrich) with TypeScript, x402 and MPP code and a "Try It Today" free-credit call to action; Breaknine is the only customer case study among recent posts. The docs list five use cases, all go-to-market: lead research, competitor brand and pricing research, people search, account-based research and news monitoring.Source

Offer task-level endpoints (for example find_work_email, search_web, enrich_company) that choose among allowed providers by a stated policy (cheapest, fastest, best success rate) within a per-call budget, fall back on failure, and report which provider answered and why. Start with one task in a deep category.

It turns Orthogonal from a catalog into the layer that makes the decision, which is far harder for rivals or single providers to copy, and fallback raises success rates for every customer.

• Providers lose volume to cheaper rivals

  Publish the policy and let quality, not only price, win.

• Outputs differ in quality

  Start where outputs are easy to compare, such as verified emails.

• #8 Publish call volume, median latency and success rate on every listing and category page

• Hacker NewsShow HN (September 1, 2026, 12 points, no comments): "one integration to discover, access, and pay for APIs" with 50+ providers and 700+ endpoints; it started with go-to-market APIs and is expanding; "The problem we're working on now is API selection", choosing providers by price, latency, reliability and result quality; the founders ask whether people would trust an agent with a budget to choose its own API provider.Source
• Your websiteSpend limits docs: a daily cap per account or organisation and one per API key, counted since midnight UTC; per-key allowed-API lists ("A call to an API outside either list is refused with HTTP 403"); refusal codes DAILY_ACCOUNT_LIMIT_EXCEEDED and DAILY_KEY_LIMIT_EXCEEDED; no alerts before a cap or a low balance is reached are described.Source
• Your websitePricing docs: "You pay only for what you use. No subscriptions."; illustrative tiers Basic $0.001 to $0.01, Standard $0.01 to $0.10, Premium $0.10 to $1.00; each response returns "priceCents"; organisation API keys bill the organisation; an agent's own account "gets a one-time $1 starter credit (one per phone number)"; wallet payments over x402, MPP or Nanopayments. The page states no platform fee and does not say whether a failed call is charged.Source
• Your websiteBreaknine case study (August 4, 2026): a tool that turns venture investors' CRM data into outreach moved its news discovery onto Orthogonal, using Fiber AI, Apollo, Serper, Bytemine, Company Enrich, PredictLeads, Fundable and Aviato; in its first two weeks it surfaced 622 updates from 233 source domains, and 539 of 622 (about 87%) were independently grounded, against "close to zero" before.Source

Become the default paid-data option in those starting points: official integrations or plugins for the most used agent frameworks, listings in MCP directories, starter templates that include an Orthogonal key with credit, and a Python SDK if call logs show demand. Co-launch each integration with the framework's maintainers.

Distribution at the point of setup compounds: every new project created from a template starts with Orthogonal connected. It positions Orthogonal as the paid-data layer next to, not against, app-connector tools.

• Maintaining many integrations

  Generate them from the OpenAPI spec and keep to the top three.

• Your websiteDocs: one MCP server URL that works with any MCP client, a one-line setup command, a CLI ("orth"), a TypeScript SDK and a REST API; guides for MCP, CLI, authentication, payments, spend limits and examples.Source
• Competitor researchRivals' homepages: Composio, "Give your AI secure access to 1,700+ apps in minutes", with a free Hobby plan of 100,000 tool calls a month and some tools billed per call; Smithery, "Connect agents to thousands of tools and services. Auth, credentials, and sessions handled for you."; OpenRouter, the model gateway Orthogonal's founders compare it to, publishes public rankings of usage by tokens processed.See the research Source
• HiringSenior Software Engineer: "one of the first engineers", full stack on Vercel with a TypeScript backend, a Next.js frontend and Postgres, payments or crypto experience a bonus. Design Engineer: builds the React/TypeScript design system and component library with care for typography, motion and accessibility, and keeps the front end fast.Source

Set budgets for LCP (under 2.5 s at p75), JavaScript per page and image weight on the homepage, catalog and provider pages; add a performance check to pull requests; and review field data monthly. Make motion respect reduced-motion settings.

Developers judge developer tools by how fast they feel, and a budget is far cheaper before a redesign than after.

• Budgets slow shipping

  Warn first, block only on large regressions.

• Site researchChrome UX Report field data for orthogonal.com (real visitors, 75th percentile): LCP 2.0 s, INP 97 ms, CLS 0.01, so the site passes Core Web Vitals, but LCP is trending 16% slower; the site runs Next.js on Vercel, with Chatwoot live chat and Cal.com booking.Source
• HiringSenior Software Engineer: "one of the first engineers", full stack on Vercel with a TypeScript backend, a Next.js frontend and Postgres, payments or crypto experience a bonus. Design Engineer: builds the React/TypeScript design system and component library with care for typography, motion and accessibility, and keeps the front end fast.Source

Let any user publish a recipe (prompt, APIs used, cost per run) on a public gallery with their tracked link. When a new user signs up through it and tops up, both get credits. Feature the best recipes monthly.

Shared recipes are useful content and a referral at once, and paying in credits keeps the cost tied to real usage.

• Credit abuse

  Pay only after a real top-up, one reward per phone number.

• #10 Turn partner spotlights into task recipes with the cost per run

• Your websiteBlog and use cases: 59 posts, the recent ones mostly partner spotlights (TinyFish on September 23, 2026, then Extruct, Akta, Financial Datasets, FullEnrich) with TypeScript, x402 and MPP code and a "Try It Today" free-credit call to action; Breaknine is the only customer case study among recent posts. The docs list five use cases, all go-to-market: lead research, competitor brand and pricing research, people search, account-based research and news monitoring.Source
• Your websitePricing docs: "You pay only for what you use. No subscriptions."; illustrative tiers Basic $0.001 to $0.01, Standard $0.01 to $0.10, Premium $0.10 to $1.00; each response returns "priceCents"; organisation API keys bill the organisation; an agent's own account "gets a one-time $1 starter credit (one per phone number)"; wallet payments over x402, MPP or Nanopayments. The page states no platform fee and does not say whether a failed call is charged.Source
• Your websiteSitemap (263 URLs, checked November 10, 2026): 85 /publisher pages, 85 /discover/<slug>/skill.md files, 59 blog posts and 18 docs pages; no /catalog/<category> pages; no /compare, /vs or /customers pages and no top-level use-case pages (only /docs/use-cases); the site has no changelog and no newsletter signup.Source