Local-first code context engine for AI agents
A drop-in alternative to Augment's context engine. Semantic code search, multi-language parsing, and MCP integration — all running on your machine.
Powerful capabilities
Semantic code search
Finds code by meaning via embeddings, not literal text
Multi-language parsing
Tree-sitter symbol extraction and analysis
Call-graph expansion
Resolves caller/callee edges with BFS expansion
Incremental indexing
Re-indexes only changed files via mtime+watcher, crash-safe
Real-time file watching
Notify, debounced, auto re-index on changes
Voyage AI embeddings
HTTP client + on-disk cache to avoid redundant API calls
LLM reranking
Reorders candidates with OpenAI/Google LLM, optional
Embedded SurrealDB
Stores chunks/symbols/edges, one datastore per repo
HTTP API + Web UI
Settings, index explorer, and query test console
MCP server
Exposes codebase-retrieval tool over streamable HTTP
SSE progress stream
Live indexing progress to UI
Large-repo scaling
Bounded memory, no O(n²) paths, built for large codebases
How it works
Index Phase
Query Phase
Supported languages
| Python | .py |
| JavaScript | .js/.jsx/.mjs/.cjs |
| TypeScript | .ts |
| TypeScript React | .tsx |
| Rust | .rs |
| Go | .go |
| Java | .java |
| C | .c |
| C++ | .cpp/.cc/.cxx/.h/.hpp/.hxx/.hh |
Other file types are chunked and embedded for semantic search only (no symbols/edges).
Install
Run with npx (recommended)
npx vibervn-context-engine@latestOr install globally
npm install -g vibervn-context-engine@latestthen run:
vibervn-context-engine --port 6699Example with custom flags
npx vibervn-context-engine@latest --port 8080 --bind 0.0.0.0Default settings
Default port 6699 with web UI at http://127.0.0.1:6699 and MCP endpoint at /mcp
Supported platforms
Linux x64/arm64, macOS arm64, Windows x64
Join us on GitHub
A drop-in alternative to Augment's context engine. 100% local-first, no cloud storage.
View Repository