Open-source coding agents
Terminal, IDE and desktop agents that read a codebase, edit files and run commands — how they gather context, apply edits and stay safe.
This category covers two kinds of tools. Coding agents run in a terminal, an IDE or a desktop app. They read your codebase, edit files and run commands for you. Prompt-to-app builders turn a description into a running web app. Three of them, vibesdk, open-lovable and bolt.diy, are in the research queue. When you choose, look at five things. How much does the agent do on its own, and how much do you approve step by step? How does it find the right code in a large repo? How are edits written, checked and undone? What stops a bad shell command, and is there a real sandbox or only an approval prompt? Which models can drive it: any text model, or only models with native tool calling?
Projects (2)
| Project | Stars |
|---|---|
| cline/clineTypeScript agent SDK behind a CLI, VS Code and desktop app: tool-calling loop, per-tool approval policies, git checkpoints, MCP and plugins. | ★ 70k |
| Aider-AI/aiderPython terminal pair-programmer: plain-text SEARCH/REPLACE edits, a tree-sitter repo map ranked by PageRank, and git auto-commits. | ★ 49k |
In the research queue: QwenLM/qwen-code, openinterpreter/openinterpreter, codewhale-hq/Codewhale, esengine/DeepSeek-Reasonix, vastsa/PI-Desktop, cloudflare/vibesdk, firecrawl/open-lovable, stackblitz-labs/bolt.diy, anomalyco/opencode, openai/codex, google-gemini/gemini-cli, OpenHands/OpenHands, aaif-goose/goose, continuedev/continue, earendil-works/pi.
Comparison questions
Each question is answered separately for every project in this category, from that project's source code.
- How is the agent loop implemented?Planner vs single loop; tool-call schema; turn structure; stop conditions; sub-agents.
- How is repository context gathered and kept within the context window?Repo maps, search/grep tools, embeddings, file-reading strategy; summarisation or compaction of long sessions.
- How are code edits applied?Edit formats (diff, search/replace, whole file, patch); how edits are validated (lint, tests, retries); undo or git integration.
- How are shell commands and file writes kept safe?Approval modes; sandboxing (containers, seatbelt, landlock); allow/deny lists; network restrictions; checkpoints.
- Which models are supported and how are they called?Providers; local models; tool calling vs text formats; per-model prompt tuning; cost tracking.
- How can it be extended and customised?MCP; plugins or custom tools; rules/instruction files (AGENTS.md etc.); hooks; headless/SDK use.