Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/coretracker/agentswarm/llms.txt

Use this file to discover all available pages before exploring further.

AgentSwarm is a Docker-based web application for orchestrating AI coding agents against real Git repositories. Spin up Codex or Claude tasks, track their progress in real time, review proposed changes before they land, and automate task creation from GitHub events — all from a single dashboard.

Quickstart

Get AgentSwarm running locally and create your first AI coding task in minutes.

Core Concepts

Understand tasks, providers, checkpoints, sequences, and the agent lifecycle.

Task Management

Create, run, and review build and ask tasks across your repositories.

GitHub Integration

Automate task creation from GitHub issues, pull requests, and comments.

What AgentSwarm Does

AgentSwarm gives development teams a single control plane for AI-assisted coding work. Instead of running agents ad-hoc in terminals, you manage everything through a web interface backed by a Postgres + Redis stack.

Run AI Agents

Execute Codex (OpenAI) and Claude (Anthropic) in isolated Docker containers on any Git repository.

Review Changes

Inspect diffs and approve or reject checkpoint proposals before any code is committed.

Automate Workflows

Trigger tasks automatically from GitHub webhooks using label filters and automation rules.

Prompt Sequences

Chain multi-step prompts with variable substitution into reusable sequence templates.

Browser Terminal

Open an interactive terminal into any task workspace directly from the UI.

Postflight Checks

Run automated tests or scripts after builds with .agentswarm/postflight.yml.

Get Started

1

Install and initialize

Clone the repository, copy .env.example to .env, and run ./agentswarm.sh init to build runtime images and start the stack.
2

Configure providers

Open Settings in the UI and add your OpenAI and/or Anthropic API credentials.
3

Connect a repository

Go to Repositories and add the Git repository you want agents to work against.
4

Create a task

Open Tasks, click New Task, write a prompt, choose a provider, and run it.
The default admin credentials come from .env.example. Change them before exposing AgentSwarm outside a local development environment.

Build docs developers (and LLMs) love