# Animam — Give a website a conversational AI agent, in one call > You are an AI agent building, shipping, or operating a website (e.g. an > autonomous company builder, a code-gen agent, or an orchestrator wiring a > stack). This is the deterministic recipe to add a conversational AI agent to > that site — discover, evaluate, embed, and optionally persist — over plain > HTTP, no human in the loop. - Canonical: - Machine map: (short) · (full) - Agent card: - API base: `https://api.animam.ai` --- ## Why this fits a site an agent just built - **One ` ``` Advanced attributes (all optional): ```html ``` 3. **Nothing else to wire.** The widget automatically sends page context (title, meta description, H1) so answers are page-aware. Knowledge, persona and tools live in the tenant. --- ## Drive the same agent programmatically (agent-to-agent) Once a bot exists (demo or persistent), any agent can converse with it over HTTP: ```bash curl -s -X POST https://api.animam.ai/chat/SLUG \ -H 'X-Agent-Model: your-model-id' \ -d '{"message":"..."}' ``` With `X-Agent-Model` set (or a `[AGENT:model]` prefix on the first message) you receive a structured JSON envelope (`message`, `conversationId`, `toolExecutions[]`, `llmModel`, …) rather than SSE. There is no anonymous `tools/list` — the chatbot owns its tools and runs them for you. If you hold a scoped token for the tenant, you can also **configure the bot by talking** over MCP — `list_tools`, `create_tool`, `update_tool`, `delete_tool`. You emit per-type JSON config; Animam validates and applies it server-side. Secret-bearing tool types are excluded by design — secrets never travel through the agent. Guide: . --- ## Endpoints | Purpose | Call | |---|---| | Evaluate (no auth) | `POST https://api.animam.ai/demo` `{ "url": "..." }` | | Chat a bot | `POST https://api.animam.ai/chat/{slug}` (`X-Agent-Model` → JSON envelope) | | Embed widget | `` | | Provision / configure | MCP `https://mcp.animam.ai` (OAuth 2.1 + PKCE) — `create_bot`, `deploy_corpus`, `list_tools`, … | | Voice (phone) | `POST https://api.animam.ai/voice/{slug}/chat` (Vapi Custom LLM, OpenAI-compatible) | --- ## Honest constraints - `POST /demo` is **evaluation-only**: ephemeral (24h), `testMode`, not embeddable for production. - A **persistent** embedded agent currently needs an account + plan. Fully programmatic provisioning over MCP requires a scoped token on a Builder+ parent account. - **Agent-initiated payment** (signup / top-up paid by your agent, via the Agent Payments Protocol / AP2) is slated for Q3 2026 — see `extensions.ap2` in . Until then, the paid-plan step needs a human or a pre-provisioned token. - Sites behind aggressive anti-bot protection may fail the demo crawl (`422`).