LeIndex
LeIndex is a semantic code search engine that can search repositories by meaning rather than relying only on matching terms.
What LeIndex is
LeIndex is a semantic code search engine that can search repositories by meaning rather than relying only on matching terms. It combines TF-IDF lexical matching with Program Dependence Graph structure and optional neural scoring from local ONNX models. Multiple repositories can be indexed and searched together. LeIndex can also run as an MCP server for AI coding tools, with listed integrations including Zed, Cursor, VS Code, Claude Code, and other tools. Its 20 MCP tools cover search, analysis, indexing, reading, editing, renaming, and impact analysis. Local ONNX models and fragment embeddings can keep data on the user's machine; OpenAI, Cohere, and compatible remote embedding endpoints are also supported, but require network access and API keys. Without an embedding backend, search falls back to offline TF-IDF, which lacks semantic understanding. Database discovery is opt-in, sensitive directories are automatically excluded, and SQL operations use parameterized queries. It is free, open source under the MIT license, and supports Linux, macOS, Windows, web, API, and self-hosted use.
Good
- Search combines lexical matching and code structure.
- Can index and search multiple repositories.
- Lists 20 MCP tools for code tasks.
- Local embeddings can keep data on-device.
- Free, open source under the MIT license.
Watch out
- Semantic search requires an embedding backend.
- Remote embedding providers require network access and API keys.
- Offline fallback lacks semantic understanding.
Who it suits
LeIndex suits developers who need to search across repositories or connect code search to MCP-compatible AI coding tools. It may also suit users who prefer local processing or self-hosting.
Verdict
LeIndex offers repository-wide code search with structural matching and optional local neural scoring. Check the embedding setup that fits your needs, since its offline fallback is keyword-based.
The facts
Each one read from LeIndex's own pages, with the date it was checked.
What it does
- AI tools
- LeIndex runs as an MCP server for AI coding tools, including Claude Code, Cursor, and other MCP-compatible agents.checked 4 Oct 2026
- Analysis
- LeIndex includes a five-phase codebase analysis pipeline and structural analysis of code relationships.checked 8 Oct 2026
- Cloud embeddings
- OpenAI, Cohere, and custom OpenAI-compatible embedding endpoints are supported and require network access and provider API keys.checked 4 Oct 2026
- Cross-project
- It can index and search multiple repositories together.checked 4 Oct 2026
- Dashboard
- An optional Bun and React dashboard provides project metrics and graph telemetry.checked 4 Oct 2026
- Deployment
- LeIndex provides Cargo, npm, PyPI, and release-script installation paths, and can run as a local CLI, MCP server, or HTTP/WebSocket server.checked 8 Oct 2026
- Embedding providers
- The README lists local Qwen3 ONNX embeddings and optional remote embeddings through OpenAI, Cohere, or a custom OpenAI-compatible endpoint.checked 8 Oct 2026
- Fallback
- Without an embedding backend, LeIndex falls back to offline TF-IDF keyword search, which the README says lacks semantic understanding.checked 4 Oct 2026
- Fallback behavior
- Without an embedding backend configured, LeIndex falls back to offline TF-IDF keyword search, which the README says lacks semantic understanding.checked 8 Oct 2026
- Install paths
- The README documents installation through Cargo, npm, PyPI, a release install script, or building from source.checked 4 Oct 2026
- Integrations
- The README documents MCP integrations for Claude Code, Cursor, VS Code, Zed, Amp, OpenCode, Qwen CLI, Gemini CLI, iFlow CLI, Droid, and Claude Desktop.checked 8 Oct 2026
- License
- The repository lists the project under the MIT license.checked 8 Oct 2026
- Local privacy
- The README says local ONNX models keep data on the user's machine, and that fragment embeddings remain local unless a remote provider is selected.checked 8 Oct 2026
- MCP tools
- The README lists 20 MCP tools for functions such as search, project mapping, reading, editing, renaming, and impact analysis.checked 8 Oct 2026
- Notable setup limit
- The README says model files are not included in release artifacts and must be downloaded by running leindex setup.checked 8 Oct 2026
- Operating systems
- The README includes Windows, macOS, and Linux installation references and runtime library formats.checked 8 Oct 2026
- Product
- LeIndex is a semantic code search engine that searches code by meaning rather than only by keywords.checked 8 Oct 2026
- Search
- It combines TF-IDF lexical matching, Program Dependence Graph structure, and optional local ONNX neural scoring.checked 4 Oct 2026
- Search methods
- Its core retrieval combines TF-IDF lexical matching with Program Dependence Graph structure, and it can add local ONNX neural scoring.checked 8 Oct 2026
- Security
- The README says database discovery is opt-in, sensitive directories such as .ssh and .aws are automatically excluded, and SQL operations use parameterized queries.checked 8 Oct 2026
- Who it is for
- The project describes itself as built for developers and AI coding tools, including use for onboarding, legacy code exploration, and cross-project search.checked 8 Oct 2026
Alternatives to LeIndex
- 1
Avidemux
Free planOpen sourceWebLinuxmacOSWindows7.3Free plan - 2
Understand
APIExtensionLinuxmacOS7.2From $100 a month - 3
ast-grep
Free planWebAPIExtensionLinux7.0Free plan - 4
Chromium
Open sourceAndroidiOSExtensionLinux6.9 - 5
Semantic Designs Source Code Search Engine
Free planAPILinuxSelf-hostedWindows6.7Free plan - 6
Code Fundi
Free planWebAPIExtension6.6From $21 a month