How to Deploy Open SWE: Step-by-Step Setup Guide

Open SWE is an asynchronous coding agent from LangChain: mention it in a Slack thread or Linear issue, and it clones your repo into an isolated cloud sandbox, works through the task, and opens a draft pull request — with minimal human oversight. Here's how to get it running for your team in five steps.

1. Requirements

• Docker with 4 GB+ RAM recommended (2 CPU / 4 GB minimum, 4 CPU / 8 GB recommended)
• A GitHub account with access to the target repos (OAuth is built in)
• A sandbox provider: Modal, Daytona, Runloop, E2B, or LangSmith
• Optional: a Slack workspace and Linear org for agent triggers

2. Deploy with Docker Compose

Save this docker-compose.yml and run docker compose up -d:

services:
  open-swe:
    image: zsssh/open-swe:latest
    restart: unless-stopped
    ports:
      - 8080:8080
    volumes:
      - ./data/open-swe:/data

3. Connect GitHub

Open the web dashboard at http://localhost:8080, sign in with GitHub OAuth, and grant access to the repositories you want the agent to work on. Use team settings to enable only allowed repos and set per-user model and profile preferences.

4. Configure triggers

Install the Slack bot and the Linear integration. From then on: comment @openswe on a Linear issue, mention the bot in a Slack thread (use repo:owner/name to pick the repo), or tag it on a PR comment to have it address review feedback on the same branch.

5. First run

Mention the agent on a simple issue. It reacts with 👀, reads the full issue context, clones the repo into a fresh sandbox, runs its curated toolset, and opens a draft PR — posting the PR link back into your thread. Send follow-up messages mid-run and it picks them up before its next step. Each task runs in parallel in its own sandbox, so your team can scale without queuing.