langchain-ai/open-swe

An Open-Source Asynchronous Coding Agent

What it solves

Open SWE provides an open-source framework for organizations to build their own internal coding agents. It solves the problem of creating secure, context-aware AI assistants that can integrate with a company's specific codebase, project management tools (Linear), and communication channels (Slack) to automate software engineering tasks with minimal human oversight.

How it works

The framework is built on LangGraph and Deep Agents, utilizing a modular architecture:

  • Agent Harness: Composes on the Deep Agents framework, allowing for customization of orchestration and tools while maintaining an upgrade path.
  • Isolated Sandboxes: Every task runs in a remote Linux environment (supporting providers like Modal, Daytona, Runloop, E2B, and LangSmith) to ensure that mistakes are contained and production systems are protected.
  • Curated Toolset: Instead of a massive library, it uses a focused set of tools for shell execution, web fetching, API calls, and interacting with Linear and Slack.
  • Context Engineering: It uses a root-level AGENTS.md file for repo-specific rules and conventions, and hydrates the agent with full issue or thread history before starting.
  • Orchestration: It supports spawning subagents for parallel subtasks and uses middleware hooks to handle real-time user feedback and error handling.
  • Invocation: Agents are triggered via mentions in Slack, Linear, or GitHub PR comments.

Who it’s for

Engineering organizations that want to deploy internal coding agents (similar to those used by Stripe or Coinbase) to automate repetitive coding tasks, fix bugs, and manage PRs within their own secure infrastructure.

Highlights

  • Multi-channel triggers: Start tasks directly from Slack, Linear, or GitHub.
  • Cloud Sandboxing: Pluggable isolated environments for safe code execution.
  • Real-time interaction: Users can message the agent mid-task, and it will incorporate the feedback before its next step.
  • Automated PR Workflow: The agent handles the entire cycle from cloning the repo and running tests to committing changes and opening a draft PR.
  • Subagent Orchestration: Ability to fan out independent tasks to child agents for increased efficiency.