{ "markdown": "
\n \n \n \n \n \"Open\n \n \n
\n\n
\n

An open-source software factory built on Deep Agents by LangChain.

\n
\n\n
\n \"License\"\n \"GitHub\n \"Built\n \"Powered\n \"Twitter\n
\n\n
\n\nOpen SWE turns engineering work into a repeatable system. Give it a code-change task from the dashboard, GitHub, Slack, or Linear—or run one on a schedule—and it works in an isolated environment to understand the codebase, make changes, validate them, and deliver a pull request.\n\nIt goes beyond code generation. Open SWE can review pull requests, learn a repository's review style, monitor CI, and respond to feedback. It is open source, deployable in your infrastructure, and designed to be adapted to your team's repositories, tools, policies, and workflows.\n\n> [!NOTE]\n> Open SWE is under active development. APIs, setup, and product surfaces may continue to evolve.\n\n---\n\n## The software factory loop\n\n```mermaid\nflowchart LR\n A[Issues, conversations, PRs, schedules] --> B[Plan and investigate]\n B --> C[Implement in an isolated sandbox]\n C --> D[Validate and deliver a PR]\n D --> E[Review, CI, and feedback]\n E -->|Follow-up work| B\n```\n\nEach cloud coding thread is bound to its own persistent sandbox, so the agent can continue from prior work when you reply. A thread is a durable conversation and work context. It can contain multiple invocations, each an agent execution triggered by a message or automation. An initial request and a follow-up belong to one thread and produce two invocations, each with its own usage. Independent threads run in parallel, and the same thread carries context from request through delivery and follow-up. Read-only PR chat does not need a sandbox, while desktop work can run directly against an allowlisted local project.\n\n## What Open SWE does\n\n### Build\n\n- Investigates repositories, plans work, edits code, and runs focused validation\n- Commits and pushes changes, then opens or updates pull requests\n- Uses subagents to parallelize research and independent work\n- Supports reusable skills, repository instructions, and custom environments\n\n### Review\n\n- Runs read-only pull request reviews on demand or automatically\n- Learns repository-specific review preferences from historical feedback\n- Supports read-only PR chat for investigating a change without modifying it\n- Keeps findings grounded in the diff and publishes them back to GitHub\n\n### Operate\n\n- Runs tasks from the web dashboard, GitHub, Slack, and Linear\n- Schedules recurring work through deterministic automations\n- Monitors opted-in pull requests with `/baby-sit`, diagnoses CI failures, and reruns only evidence-backed flaky jobs\n- Routes follow-up messages to the original thread and sandbox\n\n### Customize\n\n- Choose the models and reasoning effort available to agents and reviewers\n- Configure supported integrations and extend the curated toolset without forking Deep Agents\n- Define personal and repository coding instructions plus organization-wide review guidelines\n- Swap sandbox providers, middleware, skills, triggers, and delivery policies\n\n## API contract\n\n[`swagger.json`](swagger.json) is the generated OpenAPI 3.1 contract for the custom FastAPI backend (`agent.webapp:app`). Import it into an OpenAPI 3.1-compatible viewer, or run `make run` and open `http://localhost:8000/docs` for interactive API documentation (`/openapi.json` serves the live schema).\n\nRegenerate the file with `make swagger` after changing backend routes or models. It reflects the current route declarations: some request/response schemas and authentication requirements are not yet documented. LangGraph runtime endpoints (such as `/runs`, `/threads`, and `/assistants`) are not included.\n\n## How it works\n\n### Deep Agents is the harness\n\nOpen SWE composes the agent with [Deep Agents](https://github.com/langchain-ai/deepagents). Deep Agents provides the planning, file operations, shell access, skills, state, and subagent primitives; Open SWE adds the software-engineering tools, prompts, middleware, integrations, authorization, and product surfaces needed for end-to-end engineering work.\n\nThis composition keeps the system extensible while allowing it to inherit improvements from the underlying LangChain agent stack.\n\n### LangGraph is the runtime\n\n[LangGraph](https://github.com/langchain-ai/langgraph) provides durable execution and thread state. Each Open SWE invocation executes as a LangGraph run within a thread. Open SWE currently ships five graph entrypoints:\n\n| Graph | Role |\n|---|---|\n| **Agent** | Plans, implements, validates, and delivers software changes |\n| **Reviewer** | Performs read-only pull request reviews |\n| **Analyzer** | Learns repository-specific review style |\n| **Chat** | Answers questions about pull requests without changing code |\n| **Scheduler** | Dispatches recurring tasks and CI monitoring work |\n\n### Sandboxes contain the work\n\nCloud work runs in isolated Linux sandboxes with the development tooling supplied by the configured environment or snapshot. A sandbox persists with its thread, but an unreachable coding sandbox is not silently replaced—Open SWE fails safely rather than risk discarding uncommitted work.\n\n[LangSmith](https://smith.langchain.com/) is the default sandbox and tracing provider. Open SWE also supports [Modal](https://modal.com/), [Daytona](https://www.daytona.io/), [Runloop](https://www.runloop.ai/), [E2B](https://e2b.dev/), and local execution, with a pluggable interface for additional providers.\n\n### Tools stay curated\n\nDeep Agents supplies the core filesystem, shell, and subagent tools. Open SWE adds focused capabilities for GitHub delivery, Linear, Slack, thread management, web research, browser-based application verification, planning, review, CI monitoring, and connected services. Personal integrations load using the user's connections. Admin-configured workspace MCP tools are available to all coding-agent users.\n\n## Work where your team works\n\n- **Dashboard** — Start and continue tasks, inspect work, manage pull requests, and configure user or team settings.\n- **GitHub** — Start tasks from issues, request changes from pull request conversations, run reviews, and continue work on the same branch.\n- **Slack** — Start from a channel, thread, or code channel and receive progress and delivery updates in context.\n- **Linear** — Invoke Open SWE from an issue and post results back to the issue.\n- **Desktop (experimental)** — Run the same agent against local projects. Packaged releases currently target macOS; source builds also support Windows and Linux.\n\n## Control and safety\n\nA useful software factory needs both autonomy and boundaries. Open SWE includes:\n\n- Per-thread sandbox isolation and persistent workspaces for cloud coding tasks\n- GitHub App installation boundaries and optional per-user OAuth\n- Organization and repository allowlists with actor authorization checks\n- Credentials kept in the server process or injected through a sandbox proxy\n- Human approval before pushing workflow-file changes\n- Read-only reviewer and PR chat agents\n- Plan mode for reviewing an implementation approach before code changes\n- Opt-in automatic review and CI monitoring\n\nSandboxes can have network access and powerful tools. Deployments should use least-privilege credentials, restrict enabled repositories and integrations, and tailor approval rules to their environment.\n\n## Getting started\n\nOpen SWE includes a LangGraph backend, a web dashboard, and an experimental desktop client.\n\n- **[Installation Guide](docs/INSTALLATION.md)** — Deploy Open SWE for a team: LangGraph Platform or Docker, the GitHub and Slack apps, model providers, environment variables, and the optional Linear trigger\n- **[Development Guide](docs/DEVELOPMENT.md)** — Run it on your machine, with hot reload for the dashboard and an ngrok tunnel for webhooks\n- **[Customization Guide](docs/CUSTOMIZATION.md)** — Change models, sandboxes, tools, skills, prompts, triggers, and middleware\n- **[Open SWE Enhancement Proposals](oeps/README.md)** — Review consequential product, architecture, security, and process decisions\n\nOne deployment serves the API, the webhooks, and the dashboard from a single URL. Locally:\n\n```bash\ngit clone https://github.com/langchain-ai/open-swe.git\ncd open-swe\nuv venv\nsource .venv/bin/activate\nuv sync --all-extras\nmake build-dashboard # pnpm install + Vite build of the dashboard\nmake dev # http://localhost:2024 serves the API and the dashboard\n```\n\nCreate a GitHub App and a Slack app for your machine and fill in `.env` as described in the [development guide](docs/DEVELOPMENT.md), then sign in at `http://localhost:2024`. For UI work, `make dev-ui` starts Vite and the backend fronting it, so the same URL hot-reloads. GitHub and Slack deliver to a public webhook URL: locally the static domain of a free ngrok account (`make tunnel NGROK_DOMAIN=.ngrok-free.dev`, which exposes only `/webhooks/*`, since the dev server's LangGraph API has no authentication), on LangGraph Platform the deployment URL.\n\nProduction self-hosting uses the standalone LangGraph Agent Server and requires its license key.\n\n## Project status\n\nOpen SWE is built in the open by LangChain and is evolving quickly. The original internal coding-agent framework announcement is available on the [LangChain blog](https://blog.langchain.com/open-swe-an-open-source-framework-for-internal-coding-agents/); the project has since expanded considerably.\n\n## License\n\nOpen SWE is licensed under the [MIT License](LICENSE).\n", "bytes": 10661, "sha": "9cf4f5cb2d75a167cd56dc03c5f464acc99bbb9fef4146a1e54b2d5febc703b9", "repo_slug": "langchain-ai/open-swe", "fonte": "repo", "truncated": false, "api": "https://agentalog.com/api/listings/okf_langchain_ai_open_swe_openwiki_index_md_51254576/readme" }