HomeProjectsArticlesPK BoardAbout
Explore Projects

Open SWE

Active
GitHub Python MIT

Description

An open-source asynchronous coding agent by LangChain built on LangGraph, autonomously handling software engineering tasks including code generation, debugging, and file editing.

Key Features

  • Built on LangGraph and Deep Agents with composable architecture for easy upstream upgrades
  • Isolated cloud sandboxes (Modal, Daytona, Runloop) with full shell access per task
  • Multi-channel invocation via Slack, Linear, and GitHub with persistent thread routing
  • Subagent orchestration with parallel task fan-out and middleware hooks
  • AGENTS.md context injection for repo-level conventions and safety boundaries
  • Automatic PR creation with linter/test validation before commit

Use Cases

💡 Building internal coding agents for engineering organizations (Slackbots, CLIs, web apps)
💡 Automating software engineering tasks from Linear issues or Slack threads
💡 Running isolated code experiments and validations in cloud sandboxes
💡 Creating automated PR workflows with review feedback handling

Tags

coding-agent langgraph async software-engineering python

Categories

💻 Coding Agent

Quick Start

Install via pip: `pip install open-swe`. Set your LLM API key and LangSmith credentials. Configure sandbox provider (Modal/Daytona/Runloop). Invoke via Slack bot, Linear integration, or CLI: `open-swe --repo owner/name "fix the bug"`.
Visit GitHub

Related Projects