LLM locaux
roam-code
Local codebase intelligence CLI and MCP for coding agents.
★ 517 étoilesApache-2.0
Voir le guide →codanna est un projet de LLM locaux open source, créé par bartolli (bartolli/codanna sur GitHub). Il compte 744 étoiles sur GitHub et 70 forks, il est publié sous la licence Apache-2.0 et a été mis à jour pour la dernière fois sur GitHub le 24 septembre 2026.
Local code-intelligence MCP server and CLI for coding agents.
Get the binary per the docs.
Build the local code index.
Expose it as an MCP server.
Agents search call graphs and symbols.
Builds a local index of your repo.
Agents trace callers and callees.
Plugs into any MCP-capable agent.
Query the index from the shell.
No code leaves your machine.
Free for commercial use.
Download the source and enter the folder.
git clone https://github.com/bartolli/codanna.git
cd codannaCompile the Rust binary.
cargo build --releasePoint it at your codebase.
./target/release/codanna index /path/to/repoA local code-intelligence MCP server and CLI for AI coding agents, at bartolli/codanna.
As an MCP server exposing code search and call graphs.
No. Indexing is local-first.
Yes. Apache-2.0 licensed.
At docs.codanna.sh, linked from the repo.
Local codebase intelligence CLI and MCP for coding agents.
An MCP server plus a CLI tool that indexes local code into a graph database to provide context to AI assistants.
Code-context-engine, in Python — built for save 94% on ai coding tokens. index your cod...
Secure MCP runtime letting AI agents operate your local machines.
Open source, Git-native memory for AI coding agents - deterministic local recall
A local-first 3-in-1 AI memory layer and MCP server.
Les autres projets de la catégorie LLM locaux avec le plus d'étoiles du catalogue LibreAI — la sélection réaliste d'alternatives si codanna ne vous convient pas.
The simplest way to run open language models on your own machine.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
LLM inference in C/C++
A high-throughput and memory-efficient inference and serving engine for LLMs
🤯 LobeHub is your Chief Agent Operator, organizing your agents into 7×24 operations by hiring, scheduling, and reporting on your entire AI team.
Projets de la catégorie LLM locaux proches dans le catalogue, les plus proches en étoiles de codanna — continuez à explorer de part et d'autre.
AI gets the context.
Rustunnel, in Rust — built for self-hosted, secure tunnel server in rust. expose local.
Agent-native control plane for ComfyUI — generate media in plain chat.
ChatGPT for your home, running on a Raspberry Pi.
Check if your GPU can fine-tune an LLM — memory estimates and LoRA recipes.
Agent! now supports macOS 14.6 or later, Apple Silicon and Intel, latest release.
Gratuit et ouvert : récupérez le code source, mettez une étoile au dépôt et commencez à créer.