BrainAPI V2

For the complete BrainAPI documentation index, see llms.txt. A markdown version of any docs page is available by appending .md to its URL. Docs MCP: /docs/mcp.

BrainAPI turns text, files, and structured events into a shared knowledge base for context retrieval, ranked search, recommendations, and agent tools. You choose how much structure and ranking you need; BrainAPI keeps every request scoped to a brain.

Current documentation

These pages describe BrainAPI 2.17.0-dev. V1 remains available as Legacy, but start with V2 for new systems.

Choose your path

Build an application

Start BrainAPI, ingest one document, wait for the asynchronous task, and retrieve grounded context.

Operate BrainAPI

Deploy the services, configure providers and storage, inspect health, and diagnose failures.

Extend BrainAPI

Connect agents through MCP or add routes, lifecycle hooks, retrievers, rerankers, and recommendation models.

The product in one minute

Ingest text, files, or explicit graph facts into a named brain.
Process the input asynchronously with deterministic logic and, when configured, specialized extraction agents.
Store passages, entities, relationships, events, observations, and vectors in the configured backends.
Retrieve the representation suited to your application: context packs, ranked hits, recommendations, graph facts, or agent tools.

The brain is the isolation boundary. A request normally carries a BrainPAT token and an X-Brain-ID header; BrainAPI validates both before reading or writing data.

What you can build

Goal	Start with	Why
Ground an assistant	Context	Returns passages and graph context ready for an LLM prompt.
Search documentation or support data	Search	Returns ranked, inspectable hits with optional graph and plugin stages.
Suggest related content or next actions	Recommendations	Starts from a user or item rather than a text query.
Investigate connected events	Graph retrieval	Exposes relationships and traversal signals directly.
Give an agent tools	MCP	Lets compatible runtimes call BrainAPI without embedding REST details in prompts.

What BrainAPI is not

BrainAPI is not a replacement for ordinary CRUD when rows and foreign keys already answer the problem. It is most useful when information arrives as changing language or events, multiple representations must coexist, or a caller needs traceable relationships behind an answer.

First successful request

The Quickstart takes you from an empty machine to a completed ingest-and-retrieve flow. If BrainAPI is already running, go directly to Authentication and brains, then choose how to ingest and how to retrieve.