Structured domain model
Model a domain as doc_types, entities, and relationships so the brain has a defined structure instead of a flat prompt or ad hoc notes.
Open Index
Open Index is an open-source structured context layer for AI agents. It helps developers build searchable, domain-specific brains that agents can read, update, and serve locally or over HTTP.
Overview
Open Index is an open-source tool for building a structured context layer for AI agents. It lets you define a domain as a set of doc types, entities, and relationships, then stores and searches that data so agents can operate on something more precise than an unstructured prompt or document dump.
The project is designed around a “brain” that can be created locally, served to agents over stdio or HTTP, and kept current through manual editing or connector-driven ingestion. The default setup uses SQLite with FTS5, while OpenSearch is available for remote or multi-writer deployments that need a separate search backend.
Core capabilities
Search and retrieval
Search entities from the terminal or through the MCP layer, with support for fuzzy, typo-tolerant search in the SQLite path and native k-NN search with OpenSearch.
Dual storage modes
Keep different kinds of data in the right place by choosing file-backed or index-backed storage per doc type, which helps separate curated content from connector-fed data.
Connector ingestion and scheduled updates
Sync external data by running connectors, including MCP-based ingestion, and schedule connector runs through the built-in run command.
MCP serving and remote access
Expose the brain to agents over stdio or HTTP, with bearer-token auth available for remote HTTP serving and a config command to print the MCP block for agent setup.
Validation and explorer UI
Validate schemas, brain settings, and entity files, and provide a UI with How to use, Schema, Explore, Map, Analytics, and Jobs views for inspection.
Practical use cases
Customer support knowledge brain
A support team can model products, issues, segments, and comments so agents can search known problems and keep support knowledge current as new cases arrive.
Operations and runbook context
An infrastructure team can store services, dashboards, runbooks, and alerts in a brain that agents can query while troubleshooting or updating operational context.
Sales and account tracking
A sales or account workflow can map customers, deals, and related records so agents work from a structured pipeline rather than scattered notes.
Local agent setup
A solo developer can start a local brain in a repository, index it, and connect Claude Code or another MCP client without hosting a separate service.
Pros and Cons
Pros
- Lets you define structured domain concepts instead of relying on unstructured context alone.
- Supports both local stdio and remote HTTP serving, which covers solo use and shared team setups.
- Can keep curated, git-tracked entities in files while also supporting index-backed data owned by the search database.
- Includes validation, indexing, search, and a UI for exploring the brain after setup.
- Provides a portable Agent Skills file and Claude Code conveniences for agent setup.
Cons
- SQLite is single-writer, so remote or collaborative setups need the OpenSearch path when more than one writer will update the brain.
- The repository notes that OpenSearch is only preconfigured in the Docker path; running it outside Docker requires you to operate the cluster yourself.
- The docs show a fairly technical setup based on schemas, entity files, MCP wiring, and deployment choices, so it is aimed at users comfortable with developer tooling.
FAQ
What does Open Index do?
Open Index is a context layer for AI agents: it stores domain concepts as structured entities, searches them, and exposes them through an MCP server so agents can read and update the same brain.
How is Open Index deployed?
The repository shows a local stdio mode, a remote HTTP mode, and a Docker-based deployment path. The deployment guide says local setups are typically used for a single agent on a machine, while remote setups fit teams, cloud agents, or CI that share one brain.
Can agents both read and write context?
The docs describe read and write as the default MCP mode. Agents can retrieve context with navigation_guidelines(), search_brain(), and get_entity(), and update the brain with put_entity() and create_doc_type().
Does Open Index support Claude Code or other agent runtimes?
Yes. The docs say open-index init scaffolds a .mcp.json, CLAUDE.md, and an editing skill as conveniences for Claude Code, and the project also provides a portable Agent Skills file for other runtimes.
What search backends does Open Index support?
The project uses SQLite with FTS5 by default and can switch to OpenSearch for remote or multi-writer setups. The deployment guide notes that SQLite is single-writer, while OpenSearch is the option for several agents writing or for larger semantic search workloads.
Quick Facts
- Category
- Developer Tool
- Source domain
- github.com
- Product type
- Open-source agent context layer
- Default search backend
- SQLite + FTS5
- Remote serving
- HTTP with bearer-token auth
- Primary users
- Developers building domain-specific AI agents
Альтернативы Open Index
garden-md
garden-md is an open-source Node.js CLI that turns meeting transcripts into a local company wiki. It links entities across transcripts, keeps the original text intact, and renders the result as a browsable wiki.
Goldfish
Goldfish is an AI memory app for macOS that helps you reply, write, summarize, and continue using the context already on your Mac. It stores memory locally and writes in your tone.
Lasso
Lasso is an ecommerce product data platform for enriching catalog records, processing supplier files, generating product content, and monitoring competitors. It combines a web app with a REST API, SDK, and MCP server for teams and developers.
CreateOS Sandbox
CreateOS Sandbox is an isolated compute environment for running code and agent workloads inside Firecracker micro-VMs. It is designed for workflows that need machine-level isolation, private networking between sandboxes, and programmatic control through SDK, CLI, or MCP.
ArtDeck
ArtDeck is a visual reference board app for iPhone, iPad, and Mac that lets artists collect mixed reference material on one canvas and study it with ToolBox lenses. It works offline, supports optional iCloud sync, and is sold as a one-time purchase with no account required.
Codex Plugins
Codex Plugins bundle reusable skills, app integrations, and MCP servers into workflows you can install in the Codex app or use from Codex CLI. They help extend Codex with connected-service tasks, reusable instructions, and shared team workflows.