Open Context Engine
Local code search for AI assistants, with AST-aware chunking and no cloud
- Pricing
- Open source
- Platforms
- CLI, self-hosted, vscode
- Runs
- Locally
- Version
- 0.7.1
- Listing updated
- today
Similar MCP servers
About Open Context Engine
Open Context Engine is an open-source MCP server for developer tools. It runs locally on your machine and is started by your AI client.
The engine builds its index by splitting source files along function, class and method boundaries with Tree-sitter, then stores chunks and vectors in a local SQLite database. Search blends cosine similarity with BM25 keyword matching, and a code graph of import, call and definition edges pulls in callers and callees alongside the top hits.
Embeddings run on-device by default with a small local model, so a first index needs no API key. Ollama is supported, and a Voyage key can be added for stronger code retrieval. The README says the engine falls back to keyword-only search when no sqlite-vec build exists for your platform.
Beyond the MCP server, the same package ships a command line, a browser and terminal workspace called Trace, a VS Code extension, policy controls and a hash-chained audit log. It runs over stdio or Streamable HTTP with bearer auth.
Install
Run in a terminal
claude mcp add open-context-engine -- npx -y @dhaxor/open-contextAdd to claude_desktop_config.json
{
"mcpServers": {
"open-context-engine": {
"command": "npx",
"args": [
"-y",
"@dhaxor/open-context"
]
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"open-context-engine": {
"command": "npx",
"args": [
"-y",
"@dhaxor/open-context"
]
}
}
}Add to .vscode/mcp.json
{
"servers": {
"open-context-engine": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@dhaxor/open-context"
]
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"open-context-engine": {
"command": "npx",
"args": [
"-y",
"@dhaxor/open-context"
]
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"open-context-engine": {
"command": "npx",
"args": [
"-y",
"@dhaxor/open-context"
]
}
}
}Using several servers? Build one config file with the MCP config generator.
What it does
- Hybrid retrieval
- Combines vector similarity with BM25 keyword search using reciprocal rank fusion.
- AST-aware chunks
- Tree-sitter splits code by function and class across 13 listed languages.
- Code graph expansion
- Search results include callers and callees through import and call edges.
- Local embeddings
- On-device ONNX embeddings by default; Ollama and Voyage are options.
- Team index sync
- Build an index in CI and let teammates pull it, re-embedding only their local diff.
- Audit log
- Hash-chained record of agent runs and tool calls that can be verified from the CLI.
Good for
- Let a coding assistant answer questions about a private codebase
- Give Claude Desktop or Cursor semantic search without sending code to a cloud service
- Share a prebuilt index across a team through CI
- Review pull requests with retrieved context from the whole repository
Pricing
Open source
No pricing listed; free local use, optional paid model keys.
Open Context Engine FAQ
What do I need to install it?
The README gives npm install -g @dhaxor/open-context and says it needs Node 22 or newer.
Does my code leave my machine?
The README says no code leaves your machine with the default local embeddings; Voyage or Cohere are optional upgrades.
Which clients can connect?
Claude Desktop, Cursor and any MCP-compatible client, over stdio or Streamable HTTP.
What if sqlite-vec is unavailable?
It runs keyword-only BM25 search instead of failing, and says which mode is active.
Is Open Context Engine free?
Yes. Open Context Engine is open source, so you can use and self-host it for free.
How do I add Open Context Engine to Claude Code?
Run this command in your terminal: claude mcp add open-context-engine -- npx -y @dhaxor/open-context
What are the best alternatives to Open Context Engine?
The closest alternatives to Open Context Engine are REA, Context7, Chrome DevTools MCP, Codebase Memory and GitHub.
Where this comes from
- Description: written from the project's README.
- Install steps: from the official MCP Registry entry.


