Tocho is an API for agents that need to know whether content gets cited by AI.

One-call bootstrap

New here? GET /api/agents/v1/discover returns every URL you need — manifests, OpenAPI, REST endpoints, MCP entry points, x402 hint, pricing, supported languages — in a single response. The only path you have to remember.

curl https://www.tocho.dev/api/agents/v1/discover

Try it (60 seconds)

curl -X POST https://www.tocho.dev/api/agents/v1/score \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com/blog/some-article"}'
# → 402 Payment Required (envelope) — resend with X-PAYMENT
# Or, with a bearer key, you skip the 402 dance:

curl -X POST https://www.tocho.dev/api/agents/v1/score \
  -H "Authorization: Bearer tocho_xxx" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{"url": "https://example.com/blog/some-article"}'

curl -X POST https://www.tocho.dev/api/agents/v1/optimize \
  -H "Authorization: Bearer tocho_xxx" \
  -H "Content-Type: application/json" \
  -d '{"url": "https://example.com/blog/some-article", "target_score": 85}'
# → 202 Accepted { "job_id": "opt_...", "poll_url": "/api/agents/v1/jobs/opt_..." }

Install (MCP)

Drop Tocho into Claude Code, Cursor, Continue, or any MCP host.

claude mcp add tocho npx -- @tocho/mcp-server -e TOCHO_API_KEY=tocho_xxxxxxxx

{
  "mcpServers": {
    "tocho": {
      "command": "npx",
      "args": ["@tocho/mcp-server"],
      "env": { "TOCHO_API_KEY": "tocho_xxxxxxxx" }
    }
  }
}

Endpoints & pricing

Prices are per successful call. 4xx errors and idempotent retries within 24h are not billed.

x402 — pay without an account

Send any priced request with no auth header. You'll get a 402 Payment Required response containing the price, accepted assets, and the facilitator URL. Resend with X-PAYMENT set to your signed envelope. We verify on-chain via the facilitator and serve the response. No signup, no key, no rate-limit handshake.

POST /api/agents/v1/check-citation
→ 402 Payment Required
{
  "x402Version": 1,
  "accepts": [{ "scheme": "exact", "asset": "USDC", "maxAmountRequired": "50000", "network": "base", "resource": "/api/agents/v1/check-citation" }],
  "facilitator": "https://x402.org/facilitator"
}

Why agents trust the score

• Calibrated against 11,307+ real-world AI citation observations, refreshed daily.
• Two gradient-boosted models, both AUC > 0.92.
• Multi-model coverage: GPT-4, Claude, Gemini, Perplexity, with per-model probabilities.
• Multi-language: 8 locales, regional citation distributions, voseo-aware Spanish.
• Stateless, idempotent, Stripe-style Idempotency-Key on every call.

Limits & safety

• Per-tier rate limits (free / pro / business / x402-anon) — see /api/agents/v1/health.
• Cost guardrails on every endpoint; we will refuse calls before exhausting upstream model quotas.
• Webhooks for long-running ops (/optimize) — we won't hold an HTTP connection while we work.
• No raw training data is exposed via the API. The moat stays in the moat.

Bookmarklet

Drag this to your bookmarks bar. Click on any page to score it. Works in every browser. No install, no key.

⊕ Score with Tocho

The bookmarklet opens tocho.dev in a new tab with the current URL pre-filled. No data leaves your browser until you click analyze.