Open-source coding assistants and agents
Agents and completion tools that write, edit and run code: extensions for VS Code and JetBrains, terminal agents, a self-hosted completion server, an editor, and always-on engineering agents you run on your own infrastructure. The model each one uses is the first question, and the table says which ones can run a local model.
An MIT terminal coding agent with a full-access build agent and a read-only plan agent switched with the Tab key, installers for every platform, a beta desktop app, and the most-starred repository in this directory.
OpenAI's Apache-2.0 coding agent that runs locally in your terminal, with a desktop app and IDE extensions, signed in with a ChatGPT plan or an OpenAI API key, so the code is open and the models are OpenAI's.
Google's Apache-2.0 terminal agent for Gemini models with a free tier of 1,000 requests a day on a personal Google account, built-in search grounding, file and shell tools, MCP support, checkpointing and a GitHub Action.
A high-performance code editor from the creators of Atom and Tree-sitter, GPL-3.0-or-later with Apache-2.0 components, with an agent panel, external agents over ACP, edit prediction, MCP tools, sandboxing, and local models through Ollama or llama.cpp.
An MIT self-hosted control centre, called Agent Canvas, that runs the OpenHands agent or Claude Code, Codex, Gemini and any ACP-compatible agent across local, Docker and cloud backends, with automations triggered by schedules and webhooks.
An Apache-2.0 coding agent for VS Code, JetBrains, the terminal and a desktop app that edits files as reviewable diffs, runs commands with your approval, works with any model provider including Ollama, and connects tools through MCP.
An Apache-2.0 general-purpose AI agent from the Agentic AI Foundation at the Linux Foundation, written in Rust, with a desktop app, a CLI and an API, 15+ model providers including Ollama, and 70+ extensions over MCP.
A self-hosted coding assistant server with an Apache-2.0 core and a proprietary ee directory for the web server and team features, serving code completion and chat to VS Code, JetBrains and Vim from models on your own GPU.
Charm's terminal coding agent with 28 provider integrations, local models through llama.cpp, Ollama and LM Studio, MCP over three transports and permission prompts by default, under the Functional Source License, which becomes MIT two years after each release.
An Apache-2.0 coding agent from the Qwen team for the terminal, editor, desktop, browser and chat apps that works with OpenAI, Anthropic, Gemini and Qwen APIs or local models through Ollama and vLLM, switched at runtime, with MCP, a sandbox and subagents.
An MIT coding agent for VS Code, JetBrains and the CLI with inline autocomplete, Code, Plan, Ask, Debug and Review modes, an MCP marketplace, access to 500+ models with no API key to start, and a hosted cloud agent.
An MIT personal AI agent for the terminal that runs shell and Python locally, patches files, browses the web, works with Anthropic, OpenAI, Google, xAI, DeepSeek, OpenRouter or a local llama.cpp server, and is both an MCP client and an MCP server.
An MIT VS Code extension for fill-in-the-middle code completion and sidebar chat that defaults to a local Ollama-compatible endpoint and works offline, with nine providers and no paid tier.
Questions people ask
What is the best open-source coding assistant?
It depends on where you work. Inside VS Code or JetBrains: Cline (Apache-2.0) for an agent with reviewable diffs and any model, or Kilo Code (MIT) for autocomplete plus agent modes. In the terminal: OpenCode (MIT), Qwen Code (Apache-2.0) or Crush for a wide provider list. As an editor: Zed (GPL-3.0). For an always-on agent on your servers: OpenHands (MIT).
Which open-source coding assistants can run local models?
Cline (Ollama and LM Studio), twinny (a local Ollama-compatible endpoint by default), Tabby (a self-hosted server with its own models), Zed (llama.cpp, LM Studio, Ollama), Qwen Code (Ollama and vLLM), Crush (llama.cpp, Ollama, LM Studio), gptme (llama.cpp) and goose (Ollama). Codex CLI and Gemini CLI use only their vendor's hosted models.
Are Codex CLI and Gemini CLI really open source?
Their agent code is: both are Apache-2.0. Their models are not: Codex CLI signs in with a ChatGPT plan or OpenAI key and Gemini CLI with a Google account or key, and neither offers another provider or a local model. This directory labels the code open source and marks both as not running on your own machines.
Which one replaces GitHub Copilot's autocomplete?
Kilo Code has inline ghost-text autocomplete alongside its agent modes. twinny does fill-in-the-middle completion in VS Code from a local model. Tabby serves completion from a self-hosted server to VS Code, JetBrains and Vim. Zed has edit prediction with its own open-source Zeta model. Cline and the terminal agents do not describe autocomplete.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.