v1.3.0 is now available

The context engine your codebase deserves.

OmniContext represents a fundamental shift in AI coding. Universal dependency awareness, written in Rust, and executed flawlessly on your local machine. Supports 15 AI IDEs.

$curl -fsSL https://omnicontext.dev/install.sh | sh
Read DocsStar on GitHub
Or install the VS Code extension
Context Engine

Enterprise Context Engine.

Production-grade semantic understanding that transforms raw code into structured, queryable intelligence. Built for scale, designed for precision.

Source code
AST structure
Dependencies
Documentation
Git history
Symbols & types
File metadata
Import graph

REALTIME RAW CONTEXT

CURATED CONTEXT

Hybrid search
Context assembly
Symbol resolution
Dependency analysis
Architecture maps
Change tracking
Multi-Language Support

Tree-sitter AST parsing for 16+ languages including Python, TypeScript, Rust, Go, Java, and C++ with unified semantic extraction

Dependency-Aware

Cross-file relationship mapping with import resolution, symbol tracking, and architectural context preservation

Real-Time Updates

File system watching with incremental indexing, hash-based change detection, and sub-second propagation latency

Enterprise Scale

Connection pooling, circuit breakers, health monitoring, and graceful degradation for production deployments

Hybrid Search Engine

Best-in-class retrieval precision.

OmniContext does not rely on simple vector lookups. We fuse dense semantic vectors (usearch HNSW) with sparse exact-match keywords (SQLite FTS5) via Reciprocal Rank Fusion (RRF), ensuring your agents get exactly the context they need without hallucinating non-existent APIs.

  • usearch HNSW vector index
  • SQLite FTS5 for exact keyword matches
  • RRF fusion + heuristic boosting
Learn about Hybrid Search
omni-core hybrid-search
❯ omni search --query "auth middleware"
[1] Executing dense search (ONNX embedding)... 12ms
[2] Executing sparse search (FTS5)... 4ms
[3] Applying Reciprocal Rank Fusion...
Top Context Matches
* src/middleware/auth.rs
RRF: 0.0331
* tests/auth_integration.rs
RRF: 0.0325
* docs/api/authentication.md
RRF: 0.0150
Local-First Performance

Your code never leaves your hardware.

OmniContext is a zero-latency, local-first engine. By executing entirely on your machine with a highly parallel Rust backend and local ONNX embeddings, we ensure your code stays private and your agents stay fast—no cloud dependencies required.

  • 100% local ONNX model inference
  • Parallelized indexing (10k files < 60s)
  • Contextual chunking with semantic boundaries
  • Batch embedding with intelligent backpressure
  • Connection pooling for concurrent access
  • Encrypted SQLite index with WAL concurrency
Performance Benchmarks
omnicontext-engine — status
❯ omni status --verbose
System Telemetry
Binary RuntimeRust / Static
Memory RSS84 MB
Model Latency14ms (CPU)
Index Integrity
SQLite FTS5 (Fossilized)
HNSW Vector Index (MMAP)
Agent Protocol

Native MCP Server Integration.

OmniContext does not try to be an AI agent; it empowers the ones you already use. It runs fully locally as a standard Model Context Protocol (MCP) server over `stdio` or `sse`.

  • Provides 29 powerful MCP tools
  • Works with 15+ AI IDEs and agents
  • Stdio and HTTP SSE transports
View MCP Configuration
{ }claude_desktop_config.json
✕
123456789
{
"mcpServers": {
"omnicontext": {
"command": "omnicontext-mcp",
"args": ["--transport", "stdio", "--repo", "."],
"env": {}
}
}
}