Agents et multi-agents

Open SWE: An Open-Source Asynchronous Coding Agent by LangChain

langchain-ai/open-swe

Open SWE is an open-source framework for building internal coding agents. Built on LangGraph and Deep Agents, it provides cloud sandboxes, Slack/Linear/GitHub invocation, subagent orchestration, and automatic PR creation for engineering organizations.

★ 10,5KÉtoiles
⑂ 1,2KForks
22Problèmes ouverts
PythonLangue
MITLicence
Q@project.QualityScoreScore éditorial

Captures d'écran du projet

Capture d’écran de Open SWE: An Open-Source Asynchronous Coding Agent by LangChain Capture d’écran de Open SWE: An Open-Source Asynchronous Coding Agent by LangChain

Vue d’ensemble

Open SWE is an open-source asynchronous coding agent framework developed by LangChain. It is designed to help engineering organizations build internal coding agents—similar to those used by Stripe, Ramp, and Coinbase—that integrate with existing workflows via Slackbots, CLIs, and web apps. The framework is composed on the Deep Agents framework and LangGraph, providing an architecture that includes isolated cloud sandboxes for task execution, curated toolsets, context engineering via AGENTS.md, subagent orchestration with middleware, and multi-channel invocation. Each task runs in its own isolated cloud sandbox with full shell access, where the repository is cloned and the agent operates with contained blast radius. The agent reads issue or thread context, executes shell commands, runs linters and tests, and automatically commits changes and opens draft pull requests.

Fonctionnalités clés

  • Trigger from Linear, Slack, or GitHub
  • Instant acknowledgement with emoji reaction
  • Mid-task message injection and processing
  • Parallel task execution in isolated sandboxes
  • Built-in GitHub OAuth authentication
  • Automatic draft PR creation
  • Subagent support for parallel subtasks
  • Web dashboard for settings, team defaults, and chat UI
  • Experimental desktop app (Electron wrapper)

Prérequis, installation et démarrage rapide

Follow the installation or quick-start section in the repository README and run its dependency setup commands. Repository: https://github.com/langchain-ai/open-swe

Utilisation

Run the examples documented in the repository README and adapt the parameters to the project APIs. More usage: https://github.com/langchain-ai/open-swe

Compatibilité des modèles et cas d’usage

Not stated in the repository metadata. The repository topics include openai, anthropic, and claudecode. The README example code references a model parameter (e.g., openai:gpt-5.6-sol), indicating pluggable model support.

Licence et notes sur les risques

MIT License. License URL: https://api.github.com/licenses/mit

Editorial verification 2026-08-09: repository URL, owner, description, license and repository statistics were reviewed. License metadata: MIT. README was fetched for the channel draft; re-check repository dependencies, releases and model terms before production use.

Publication et maintenance

Not stated in the repository metadata. Created on 2025-05-21. Last pushed on 2026-08-05.

Firecrawl

firecrawl/firecrawl

★ 161,1KTypeScript

LangChain

langchain-ai/langchain

★ 143,6KPython

RAGFlow

infiniflow/ragflow

★ 86,7KGo