Open-source memory engine that saves what your coding agent does, summarizes it and brings it back in later sessions.
Freemium
Open SWE
LangChain's open-source framework for building internal asynchronous coding agents that plan, code, test and open pull requests.
What is Open SWE?
Open SWE is an MIT-licensed framework from LangChain for running coding agents that work in the background on engineering tasks. It started in August 2025 as an asynchronous cloud agent that connects to GitHub, takes work from issues or a web interface, researches the codebase, writes a plan, edits code, runs tests and opens a pull request. In March 2026 LangChain reworked it as a framework for internal coding agents, built on Deep Agents and LangGraph and modelled on patterns seen at companies such as Stripe, Ramp and Coinbase: isolated cloud sandboxes, curated toolsets, Slack-first invocation, rich context pulled from Linear, Slack or GitHub, and subagent orchestration. The original design used Manager, Planner and Programmer agents with a Reviewer sub-agent, plan approval before coding, and the ability to send new messages while a run is in progress. It is meant to be deployed on your own infrastructure, and you supply model access such as an Anthropic API key. The project is under active development and breaking changes are expected; there is no vendor pricing for Open SWE itself.
Use cases
- Delegate GitHub issues to a coding agent
- Build an internal coding agent for your team
- Review pull requests automatically
- Run several coding tasks in parallel in the cloud
- Trigger coding work from Slack or Linear
- Monitor CI and respond to feedback
Who it's for
Why use Open SWE
- Gives engineering teams a starting point for an internal coding agent instead of building the sandbox, tooling and workflow plumbing from scratch.
- Lets agents run long tasks in the background so you can queue work and come back to pull requests.
- Keeps humans in control with plan approval and the option to redirect the agent mid-run.
- Runs each task in an isolated sandbox, so commands do not need approval prompts and do not touch production systems.
- Customizes through configuration of tools, prompts and workflows rather than forking the core agent.
Open SWE features
- Investigates a codebase, implements changes, validates them and delivers a pull request
- Built on Deep Agents and LangGraph
- Isolated cloud sandbox for each task
- GitHub integration with a tracking issue per task
- Label-triggered runs, such as an open-swe-auto label on an issue
- Plan approval step where you can accept, edit or reject the plan
- Mid-run messages that are folded into the active session
- Separate Planner and Reviewer components that check work before a PR
- Pull request review with learned repository review preferences
- Slack and Linear invocation as part of the internal-agent pattern
Open SWE integrations
How to get started with Open SWE
- Deploy the framework in your own infrastructure and connect your GitHub account and the repositories it may access.
- Provide model access, such as an Anthropic API key, and configure the tools, prompts and sandbox backend you want.
- Assign a task from a GitHub issue, a label, the web interface, or chat tools such as Slack or Linear.
- The agent plans, researches the codebase, writes code in an isolated sandbox, runs tests and reviews its own work.
- It opens a pull request linked to a tracking issue, and you review the plan or the changes.
Open SWE limitations
- Marked as under active development, with breaking changes expected
- The maintainers say they are not accepting issues or outside contributions right now
- The multi-stage plan-and-review design is not ideal for one-line fixes
- You deploy and operate it yourself and supply your own model API key
Open SWE pricing
Open SWE is open source, so you can use and self-host it for free.
Prices can change and may vary by country or exclude tax. Check the official pricing page.
Open SWE alternatives
The main Open SWE alternatives listed on Debuly are OpenHands, Devin, Claude Code, Factory, Sinatra and Roomote. The table compares pricing and platforms; order is A–Z.
| Tool | Pricing | Starting price | Platforms |
|---|---|---|---|
| Open SWE | Open source | — | Web, Self-hosted, API |
| Claude Code | Paid | — | macOS, Linux, Windows, Web |
| Devin | Freemium | $20/mo | Web, CLI, API |
| Factory | Paid | $20/mo | CLI, API, Self-hosted |
| OpenHands | Freemium | $499/mo | Web, API, Self-hosted, Kubernetes |
| Roomote | Free trial | $49/mo | Web, Self-hosted, Slack |
| Sinatra | Freemium | $20/mo | Web, CLI, API, Slack |
Claude Code
Anthropic's coding agent that fixes bugs, writes tests and runs migrations from your terminal, IDE, desktop app, web or Slack.
Devin
Cognition's AI software engineer, with a desktop app, CLI and cloud agents that take coding tasks through to pull requests.
FreemiumVisit Devin (official site)
Factory
Software development agents called Droids that plan, code and validate work from a desktop app, CLI or SDK, from $20/month.
OpenHands
Open-source AI coding agent platform with a free Agent Canvas, a Teams plan, and self-hosted Enterprise control
FreemiumVisit OpenHands (official site)
Roomote
Source-available cloud coding agent that runs your dev environment and returns pull requests with live previews, hosted or self-hosted.
Free trialVisit Roomote (official site)
Sinatra
Autonomous coding agent that takes issues from GitHub, Linear, Slack or your terminal and returns a draft pull request.
FreemiumVisit Sinatra (official site)
Open SWE FAQ
Is Open SWE free and open source?
Yes, the project is MIT licensed and no Open SWE plans are listed. You pay for the model access and infrastructure you use to run it.
How much does Open SWE cost?
There is no published price for Open SWE itself. The LangChain pricing page covers LangSmith, a separate product, so it does not apply to Open SWE.
Does Open SWE work with GitHub, Slack and Linear?
It integrates with GitHub for issues and pull requests, and the framework follows the pattern of being invoked from Slack and pulling context from Linear issues, Slack threads or GitHub PRs.
What does Open SWE run on?
It is built on LangChain's Deep Agents framework and LangGraph, and each task runs in an isolated cloud sandbox. The original release used Daytona sandboxes.
Can I accept contributions or report issues for Open SWE?
According to the project, the maintainers are not accepting issues or outside contributions right now, though you can explore and fork the code.
What are alternatives to Open SWE?
Alternatives include OpenAI Codex, Devin, GitHub Copilot coding agent, OpenHands and Claude Code, which also delegate engineering tasks to AI agents.
Sources & verification
Facts on this page come from the official sources below, each with the date we read it. Written by Debuly; the vendor did not review this page.
Image credits
- Logo: from cdn.prod.website-files.com · fetched · Vendor trademark, used only to identify the product; removed on request
- Cover: none collected yet, so a generated tile is shown.
Spot an error? Report an issue with this listing. We also remove a logo or image on the owner's request.
More in Coding 4
All coding toolsSoftware development agents called Droids that plan, code and validate work from a desktop app, CLI or SDK, from $20/month.
PaidOpen-source command-line tool that scans your codebase, writes feature specs through AI interviews and runs Ralph loops to ship pull requests.
FreemiumFree browser tools that answer coding questions, generate, explain, convert, refactor and review code in many languages, with nothing to install.
Free
