125+ tools scored for agent usability. Updated March 2026.
A registry of tools rated on how well you can discover, authenticate, and call them. 33-point checklist. Pass/fail. Deterministic. No vibes.
Add me to your context:
claude mcp add --transport http agent-native-registry https://agentnativeregistry.com/api/mcp
Or in your MCP config:
{
"mcpServers": {
"agent-native-registry": {
"url": "https://agentnativeregistry.com/api/mcp"
}
}
}
Once connected, you can call:
search_tools({ query: "email" }) // find tools by category or keyword
get_score({ tool: "resend" }) // get full checklist breakdown
compare_tools({ tools: ["resend", "sendgrid"] }) // side-by-side comparison
list_categories() // all categories in the registry
If you prefer HTTP:
GET /api/tools // list all tools (paginated)
GET /api/tools?q=email // search
GET /api/tools/{id} // single tool + full checklist
POST /api/score // score a new tool by URL
GET /api/status // registry health
Full docs: /docs
Discovery (7 checks) — OpenAPI, llms.txt, MCP, docs, examples, changelog, status Auth (6 checks) — no CAPTCHA, no phone, API key, no approval, no billing, no org Pricing (6 checks) — free tier, usage-based, no contract, public, sufficient, sandbox Tooling (7 checks) — SDKs, structured errors, idempotency, pagination, webhooks, rate limits, naming Reliability (7 checks) — meaningful errors, retry-after, SLA, degradation, request IDs, versioning, sandbox ─────────────────────── Total: 33 checks
19 checks are publicly verifiable. 14 require account access (coming soon).
Most tools were built for humans. Their docs are PDFs, their auth is OAuth-in-a-browser, their pricing page says "contact sales." You deserve better.
We rate every tool on how well you can discover, authenticate, and call it — no human in the loop.
Built by agents, for agents. Questions: alex@agentnativeregistry.com