Open Index MCP
An MCP server that keeps domain knowledge in a graph: the agent searches entities and relations and can add or update them itself.
- 114 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- MCP writes entities to the database by default; HTTP mode exposes that database to remote agents by token.
Install
- Works in
Manual install
pip install -e '.[all]'First README Quickstart command: installs the core, explorer UI and MCP server from a local clone. `.[mcp]` is enough for MCP alone.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- MCP writes entities to the database by default; HTTP mode exposes that database to remote agents by token.
This is third-party code. Review the repository files before installing.
What it does
Open Index builds a brain for a domain: a set of document types, entities and relations between them. Types are described by YAML schemas, and entities live in JSON files or in the search index. Data is stored in SQLite with full-text search by default, and an OpenSearch backend exists for larger volumes. The `open-index mcp` command starts an MCP server through which the agent calls `search_brain`, `get_entity` and `navigation_guidelines`. For writing there are `put_entity`, `put_entities` and `create_doc_type`. The `--read-only` mode disables writes. For remote agents `open-index serve` exposes the same interface over HTTP with a token.
Who it is for. Support, sales and infrastructure teams building an agent for their domain who want to give it a structured knowledge base instead of scattered documents.
Good fit when
- A support agent must know products, common issues and the links between them.
- Descriptions of services, runbooks and dashboards must live in one graph.
- Several agents query a shared knowledge base over HTTP.
- Knowledge has to be loaded from CSV, JSONL or another MCP server.
Not a fit when
- You need plain search over a few notes; a graph with types and schemas is excess work.
- Data must stay unchanged and `--read-only` is not enabled.
- The environment has no Python 3.10.
Example request
Search the graph for known issues of the payment service, show linked runbooks and add a new entity for the latest incident.MCP
- Transport
- stdio, http
- Authentication
- not required
Limitations
Python 3.10 or newer is required. The `pip install -e '.[all]'` command runs from a local copy of the repository. MCP is read and write by default, so the agent can change entities; enable `--read-only` for a safe mode. Entities with `index` storage live only in `brain.db`, which git ignores, so back it up separately. The project is young, package version 0.1.0, and the latest commit dates from August 2026.
How to disable. Remove the `open-index` entry from `.mcp.json` or the client configuration and stop `open-index serve`. The brain directory and `brain.db` are deleted separately.
FAQ
How does an entity differ from a document type?
A type (`doc_type`) describes a concept and its fields, such as `service`. An entity (`entity`) is a concrete instance of that type with `related_to` links.
How do I stop the agent from changing data?
Run `open-index mcp --read-only` or `open-index serve --read-only`.
How do I connect a remote server?
Run `open-index serve --port --token`, then `open-index mcp-config`; it prints a connection block for the client.
Where is data stored by default?
In SQLite with FTS5 full-text search. An OpenSearch backend is available for bulk data.
Is there a ready example to try?
Yes, `examples/support-brain`: run `open-index index --brain examples/support-brain`, then `open-index ui` to browse the map.
Discussion
No questions or reviews yet.
Related
MCP serversAn MCP server and REST API with shared memory for Claude Code, Codex and Gemini: Obsidian notes, sessions, bug fixes, lean search.
Self-hosted long-term memory for agents: semantic search, a knowledge graph and autonomous consolidation over MCP and a REST API
A CLI and MCP integration keeping agent decisions and checkpoints in local SQLite so work survives a session change.
BrainAPI turns text into a knowledge graph on your server and serves agents over MCP, REST and a console with fact-path answers
An MCP server and API of shared memory for several agents: facts, decisions and events live in one database with semantic search.
Shared memory for fleets of AI agents with access scopes, an audit trail and a knowledge graph: an MCP server, Docker and a cloud
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Open Index (this card) | no | not required | 114 | |
| Create Context Graph | no | 747 | generates a context graph for an app; here a persistent brain with types, search and writes over MCP | |
| MCP Memory Keeper | no | not required | 136 | keeps session memory; here a structured domain knowledge base with schemas and relations |
README in short
The README sets out the four brain primitives, a quick start with a support example, a table of commands and the directory layout. Separate sections cover MCP, storage and cloud deployment.