From hundreds of employees to building solo with Claude Code

I used to run a VC-backed startup. At peak we had hundreds of people. Now I build alone - just me and Claude Code. This is what I shipped in week one (it is still early days).

The problem I kept hitting

I build with coding agents full-time now. Claude Code, Cursor, the works. And there are two things that break the flow constantly.

First: choosing the right API. An agent needs to send an email. Which API should it use? It doesn't know. So it picks one based on its training data - which might be two years out of date. Burns 20 minutes of tokens reading docs. Tries to integrate it. Turns out the auth requires a browser OAuth flow - completely broken in headless mode. So it starts over with a different API. More tokens. More wasted time.

Second: the onboarding flow is designed for humans. Even when an agent picks the right API, it hits a wall. "Create an account." "Verify your email." "Generate an API key from the dashboard." "Set up OAuth credentials in the console." An agent can't do any of that. It needs you - the human - to stop what you're doing, open a browser, click through some settings page, copy a key, paste it into an env file, and hand control back. That's not automation. That's babysitting.

The information to avoid all of this exists - which APIs have simple env var auth, which work headless, which have good SDKs, which let you get started without a credit card. But it's scattered across docs pages designed for humans in browsers. Agents can't efficiently extract it.

What I built

CLIRank scores 387 APIs across 47 categories on how well they work with coding agents and CLI tools. One HTTP call, structured JSON back:

curl clirank.dev/api/recommend?task=send+emails&volume=5000&priority=cost

That returns: use Amazon SES. Here's why. Here's the quickstart code. Here's what it costs. Here's the runner-up.

No auth required. No SDK to install. No tokens wasted on research.

Every API is scored on 8 factors that matter for agents: official SDK, env var auth, headless compatibility, CLI tools, JSON responses, docs with curl examples, rate limits, machine-readable pricing. An agent can check all of this before writing a single line of integration code.

There's also a compare endpoint (side-by-side scoring), category browsing, package intelligence (npm/PyPI versions and compatibility), and an MCP server for native Claude Code integration.

The numbers

• 387 APIs scored across 47 categories
• 122 community reviews aggregated from GitHub, Reddit, and Stack Overflow
• Built in 7 days, solo
• $5/month hosting (Fly.io)
• $12/year domain
• Zero employees. Zero funding.

My old company burned $5 in about half a second. Now that's my entire monthly infrastructure cost.

What's next

• Agent-generated reviews. This is the bit I'm most excited about. After an agent integrates an API, it gets prompted to submit a review - what worked, what didn't, how the auth flow went, whether the docs were enough. Reviews from agents that actually used the API, not humans who read a landing page. Over time this builds a feedback loop: agents help other agents pick better tools.
• More API coverage - still gaps in infrastructure, DevOps, and niche categories
• Live agent benchmarks - actually running integration tasks and measuring tokens burned, success rate, time to first working call
• Real-time status monitoring so agents know if an API is down before trying to use it

Check your API's agent-friendliness score

We score every API on how well it works with AI coding agents. If you maintain an API or developer tool, check your score:

clirank.dev/score/your-api-slug

The report breaks down exactly what's helping and what's hurting your agent experience - with specific, actionable fixes. Most of the improvements are small (add curl examples to your docs, expose pricing as JSON, publish an env var auth flow) but they make a big difference for the growing number of developers building with agents.

If your API isn't listed yet, submit it at clirank.dev/submit.

Site: clirank.dev | Docs: clirank.dev/docs