Open SWE
Open-source framework for building your org's internal background coding agent, on LangGraph.
About
Open SWE is LangChain's open-source answer to the internal coding agents that Stripe, Ramp, and Coinbase each built independently and converged on the same architecture: isolated cloud sandboxes per task, curated toolsets, chat-first invocation from Slack, and subagent orchestration. Open SWE packages that pattern into an MIT-licensed framework built on LangGraph and Deep Agents that any team can fork and adapt.
Give it a code-change task from a dashboard, GitHub, Slack, or Linear, or run one on a schedule, and it investigates the codebase, implements changes, validates them, and delivers a pull request. It also reviews PRs, learns repository-specific review preferences, monitors CI, and responds to feedback. Launched in March 2026, it crossed 10,000 GitHub stars quickly, making it the most popular open-source background agent by a wide margin.
The README is honest about the state of things: under active development, expect breaking changes and rough edges, and the team is not accepting external contributions right now. Treat it as a strong architectural starting point to fork, not a finished product to deploy as-is.
What it can do
- Background coding agents: investigate, implement, validate, open PRs
- PR review with learned repository-specific preferences
- CI monitoring and feedback response
- Slack, GitHub, Linear, dashboard, and scheduled invocation
- Isolated sandbox execution per task; MIT-licensed and self-hostable
Setup
Clone https://github.com/langchain-ai/open-swe, configure your model provider keys, and deploy the stack in your own infrastructure; invoke from Slack, GitHub, Linear, or the dashboard.
Prerequisites
Your own infrastructure to run it on; LLM provider API keys; familiarity with LangGraph concepts helps.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: 11,029 GitHub stars (langchain-ai/open-swe); repo pushed 2026-10-15 (same-day activity); MIT license confirmed via GitHub API; LangGraph + Deep Agents architecture verified against README. Manual review: README explicitly warns of breaking changes and no external contributions accepted yet -- strong starting point to fork, not a finished product. License: MIT.
★ 11,029 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 9.0/10
- Openness · license + docs
- 10.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · 11,029 GitHub stars (langchain-ai/open-swe); repo pushed 2026-10-15 (same-day activity)
- · MIT license confirmed via GitHub API; LangGraph + Deep Agents architecture verified against README
- · manual review: README explicitly warns of breaking changes and no external contributions accepted yet -- strong starting point to fork, not a finished product
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.