Pre-release v0.1.0

Introduction

AgentFork is infrastructure that makes open-source repositories instantly forkable. When someone forks your project, they get more than just code — they get a fully provisioned environment with databases, services, environment variables, and a live preview URL.

No Docker. No local setup. No "works on my machine." Whether it's a first-time contributor, a senior engineer evaluating your architecture, or an AI agent writing a patch — everyone gets a working environment in seconds.

AgentFork is currently accepting waitlist signups. We're onboarding projects in batches. Join the waitlist to start using it with your repositories.

Quick Start

Get a repository running on AgentFork in three steps.

2

Select a repo and hit Index

Pick a repository from your account and press Index. AgentFork's AI will scan the entire codebase — config files, dependency manifests, ORM schemas, Docker definitions — and automatically generate an agentfork.yml blueprint with everything detected.

auto-generated agentfork.yml
# auto-detected by AgentFork AI
name: my-project
runtime: node-20
framework: sveltekit
services:
  postgres: "15"
  redis: "7"
setup:
  - bun install
  - bun prisma generate
  - bun prisma db push
start: bun run dev
3

Go live

That's it. Your project is now forkable. Every fork automatically gets a fully provisioned environment — databases, services, environment variables — and a unique live preview URL. Contributors go from fork to running code in seconds.

Stack Detection

How AgentFork understands your project.

When you index a repository, AgentFork's AI engine scans the entire project and builds a dependency graph. It reads config files, dependency manifests, ORM schemas, Docker definitions, and environment files to determine exactly what your project needs to run — including all environment variable names.

Everything detected is written into an agentfork.yml file. You can review it, customize it, or just let the AI handle it entirely.

What gets detected

SignalFilesDetects
Package managerpackage.json, bun.lockbRuntime, dependencies, scripts
Framework configsvelte.config.js, next.config.*Framework, build pipeline
ORM schemasschema.prisma, drizzle.config.*Database type, migrations
Dockerdocker-compose.ymlServices, ports, volumes
Environment.env, .env.example, .env.localAll env variable names, connection strings

Environments

Isolated, ephemeral infrastructure for every fork.

Each fork of your project gets a completely isolated environment — its own database instances, cache layers, environment variables, and compute. Environments are provisioned on-demand when a fork is created and automatically torn down after a period of inactivity.

Lifecycle

1
Provisioning — Fork detected. Databases created, migrations run, seed data loaded, dev server started.
2
Active — Environment is live. Code changes hot-reload. Preview URL is publicly accessible.
3
Sleeping — No activity detected. Environment suspended to save resources. Reactivates instantly on next visit.
4
Terminated — Extended inactivity. Environment and data permanently removed.

Live Preview URLs

Every fork gets a unique, shareable URL.

When an environment is provisioned, AgentFork assigns it a unique subdomain on *.agentfork.dev. These URLs are publicly accessible and can be shared in pull requests, issue comments, or Slack threads.

URL format

Pattern
https://{fork-id}.agentfork.dev
ContextExample URL
Pull requestpr-42.agentfork.dev
Branch previewfeat-auth-redesign.agentfork.dev
Contributor forkjdoe-my-project.agentfork.dev

Supported Stacks

Frameworks, runtimes, and services we support today.

Frameworks

  • Next.js
  • SvelteKit
  • Nuxt
  • Remix
  • Express / Fastify
  • Rails planned
  • Django planned

Services

  • PostgreSQL
  • Redis
  • MySQL
  • MongoDB
  • Elasticsearch planned
  • RabbitMQ planned
  • MinIO (S3) planned

agentfork.yml

The configuration file that powers every environment.

The agentfork.yml is the single source of truth for how your project's environments are provisioned. It's auto-generated when you index a repository — the AI detects your framework, runtime, databases, setup commands, environment variables, and dev server configuration. You can customize it anytime from the dashboard.

Full reference

agentfork.yml
# Project name (defaults to repo name)
name: my-project
# Runtime and framework
runtime: node-20
framework: sveltekit
# Services to provision
services:
  postgres:
    version: "15"
    extensions: [pgvector, uuid-ossp]
  redis:
    version: "7"
# Commands to run during setup
setup:
  - bun install
  - bun prisma generate
  - bun prisma db push
  - bun run seed
# Dev server command
start: bun run dev
# Port to expose (auto-detected if omitted)
port: 5173

You don't need to write this yourself. The AI generates the entire agentfork.yml when you index your repo — including environment variable names, setup commands, and service versions. You can edit it from the dashboard at any time.

API Reference

Programmatic access to AgentFork.

API documentation is coming soon.

The REST API will let you programmatically manage environments, trigger indexing, and query status. Join early access to be notified when it launches.

Ready to make your repo forkable?

Get early access and start onboarding your project today.

Get Early Access