# Agent Seek developers

Scoped index for self-serve onboarding. Full site index: [/llms.txt](/llms.txt). There is no contact-sales form.

Install the skill first: [/.well-known/agent-skills/agent-seek/SKILL.md](/.well-known/agent-skills/agent-seek/SKILL.md) — then [Install the skill](/llms.txt#install-the-skill) → [/auth.md](/auth.md) → MCP or `POST /v1/search`.

## Get credentials without a human

1. **Zero-auth inspect:** `GET /v1/sandbox` (canned SearchResponse), `GET /health`, `GET /v1`, `GET /openapi.json`. Live host does not mint self-serve API keys.
2. **Anonymous OAuth (self-serve):** `POST /agent/identity` with `{"type":"anonymous"}`, then exchange the `identity_assertion` at `POST /oauth2/token` (`grant_type=urn:ietf:params:oauth:grant-type:jwt-bearer`). Yields `health:read` + `mcp:invoke`. Claim later for `search:read`.
3. **Dynamic client registration:** `POST /oauth2/register`, then authorization code + PKCE S256 at `/oauth2/authorize`.
4. **Local sandbox key:** `cp .env.example .env` and set `AGENT_SEEK_API_KEY`. Default example value is `dev-agent-seek-key-change-me`. Never commit it.

Live prototype keys are still operator-issued if you need `search:read` on https://agentseek.dev without claiming. Email is optional, not a gate for local or anonymous use.

## Developer pages

- [Agent Seek API](/developers) · [developers.md](/developers.md)
- [auth.md](/auth.md)
- [pricing.md](/pricing.md)
- [OAuth AS metadata](/.well-known/oauth-authorization-server)
- [Protected resource](/.well-known/oauth-protected-resource)
- [Skill](/skills/agent-seek/SKILL.md)

Default `mode=deep` (Stage A survivor fetch, cap 12; not a full-web crawl). Pass `mode=snip` for cheaper title/URL/snippet ranking. Signals include `prompt_injection` (UI: **Injection risk**). Hard gates (order): subject_match → is_republisher → prompt_injection (0.55, fail-open). `k` default 10; `max_candidates` default 50 (hard max 100).

Rate limit: 60 requests / minute. Errors: `application/problem+json`.