# GTMRouter — the unified GTM API for AI agents

> One API key + one prepaid credit balance runs the full outbound loop:
> find leads, check intent, enrich to verified emails, verify deliverability,
> provision warmed sending, draft campaigns, launch (human-approved), read replies.
> LLM passthrough included. Misses are always free; every response discloses
> its exact charge in the `gtmrouter` envelope.

Base URL: https://api.gtmrouter.dev
MCP server: https://mcp.gtmrouter.dev/mcp  (Authorization: Bearer <your key>)
Console (accounts, keys, approvals, top-ups): https://app.gtmrouter.dev
OpenAPI spec: https://api.gtmrouter.dev/openapi.json

## Verbs (REST paths; also exposed as MCP tools of the same names)
- POST /v1/find_leads — search 200M+ B2B contacts by titles/industries/locations/headcount. 0.2 credits per returned lead. Emails come back masked; chain into enrich.
- POST /v1/enrich — reveal verified work email (+ mobile) for a person (name+domain, linkedin_url, or email). 5 credits on hit only; 90-day workspace cache.
- POST /v1/verify_email — deliverability with catch-all analysis. Statuses: valid | invalid | catch_all_valid | catch_all_unknown | unknown | risky. 1 credit.
- POST /v1/find_intent — hiring/funding/news/job_change/technology signals for up to 100 company domains. 10 credits per signal-bearing company + 0.5 credits per company scanned.
- POST /v1/setup_sending — quote (free) → human console approval → purchase warmed sending capacity. 1,500 credits/domain one-time + 750 credits/mailbox/month upkeep. Requires assets:purchase scope + Idempotency-Key. Returns operation_id; poll /v1/operations/:id.
- POST /v1/campaigns — create a DRAFT campaign (never sends). Free. Suppressed/invalid leads filtered and reported. Requires Idempotency-Key.
- POST /v1/launch_campaign — preview → human console approval → launch. Free. Requires campaign:launch scope, Idempotency-Key, and active sending capacity. Suppression re-checked at launch; fails closed.
- POST /v1/get_replies — read replies (classifications: interested, not_interested, out_of_office, unsubscribe, bounce, other). Free. Repliers auto-suppressed.
- POST /v1/llm/chat — OpenAI-compatible chat completions. Models: anthropic/claude-sonnet-5 (0.5 cr/1k tok), anthropic/claude-haiku-4.5 (0.25), openai/gpt-5-mini (0.1).
- GET /v1/usage, /v1/requests/:id, /v1/operations/:id — balance, request status, async progress. Free.

## Rules an agent must know
- Auth: `Authorization: Bearer gtmr_sk_…`. Scopes gate verbs; purchases/launches additionally need a single-use confirmation_token approved by a human in the console. Agents cannot spend or send alone.
- Idempotency: send an Idempotency-Key header; retries replay the original result without re-charging. Required on setup_sending purchases, create_campaign, launch_campaign.
- Suppression is permanent and server-enforced: replies/unsubscribes/bounces are never contacted again.
- Errors: single envelope {error:{code,message,request_id?,details?}}. Retry only rate_limited, provider_unavailable, internal (same Idempotency-Key). operation_unknown_state → poll, never blind-retry.
- Free tier: 300 credits at signup, no card. 1 credit ≈ $0.01. Rate limit 300 req/min/key.

## Docs
- Quickstart: https://docs.gtmrouter.dev/
- Add to your agent (MCP): https://docs.gtmrouter.dev/agent.html
- Concepts: https://docs.gtmrouter.dev/concepts.html
- Pricing: https://docs.gtmrouter.dev/pricing.html
- API reference: https://docs.gtmrouter.dev/reference.html
- Errors: https://docs.gtmrouter.dev/errors.html
- Playbooks: https://docs.gtmrouter.dev/playbooks.html