Your codebase.
AI's cloud workspace.

An isolated environment with your repository cloned, AI coding agent installed, and full git workflow ready. The agent reads, edits, commits, and pushes — even while your laptop is off.

Start for free → See how it works
4
AI Agent Harnesses
3
Git Providers
11
Dev Integrations
∞
Persistent Sessions
How it works

Three steps to a working session

Connect your repo, pick a branch, chat with AI. The agent works in a real environment with real tools — not a sandbox.

1

Connect your repo

Link GitHub, GitLab, or Bitbucket — including self-hosted instances. Personal access tokens with configurable scopes. Your repos appear instantly on the dashboard.

2

Environment boots up

An isolated environment launches with your repo cloned, git configured, Python 3.12, Node.js 22, and your chosen AI coding agent pre-installed. Ready in seconds.

3

Chat and ship

The agent reads files, edits code, runs tests, commits changes, and creates merge requests. Close your browser — the container keeps running. Resume anytime with full state.

Features

Not just a chat. A full workspace.

Every session runs in its own isolated environment — a real development environment with full access to your codebase, git, and dev tools.

Always-on cloud containers

Containers persist in the cloud. Shut your laptop, close the tab. Come back hours later — your workspace, diffs, and agent context are all restored exactly where you left off.

Full git workflow

The agent commits and pushes directly. Create branches, open merge requests, resolve review threads. GitHub CLI (gh), GitLab CLI (glab), and Bitbucket CLI (bkt) are pre-configured.

Complete isolation

Every session gets its own isolated environment. No conflicts between experiments, no worktrees, no local environment setup. Sessions stay separated with network sandboxing and restricted capabilities.

Scheduled automation

Cron jobs run prompts on a schedule: nightly code reviews, recurring refactoring, test sweeps. Sessions are created automatically and you get notified when done.

Real-time file preview

Watch file changes as the agent works. See diffs, browse the file tree, and preview modifications before they're committed. Select code to ask questions or request refactoring.

External API & Webhooks

Create sessions programmatically via REST API. Configure webhooks to receive session.completed and session.error events with HMAC-SHA256 signed payloads.

Issue & MR boards

Browse issues and merge requests directly from the dashboard. One-click auto-solve: the agent reads the issue, analyzes the code, implements a fix, and creates a merge request. Works in bulk.

Skills & configs per harness

Define custom skills, agents, hooks, scripts, extensions, and rules per harness. Enable per-repository. Injected into containers on session start.

Bring your own key

Use your own Anthropic or OpenAI API key, or connect via Claude/ChatGPT subscription (OAuth). Switch models and thinking effort per session. Full cost tracking with usage analytics.

Agent harnesses

Four AI coding agents. Your choice.

Each harness runs natively in the container with its own CLI, session persistence, and configuration format. Switch between agents per session.

Claude Code

Anthropic's official CLI agent with deep codebase understanding and extended thinking.

Opus 4.6Sonnet 4.6Haiku 4.5

OpenAI Codex

OpenAI's code-focused agent with thread persistence and full codebase access.

GPT-5.4GPT-5.4 MiniGPT-5.3 Codex

Pi.dev Agent

Multi-provider agent supporting both Anthropic and OpenAI models with extensions.

Claude SonnetClaude OpusGPT-5.4

OpenCode

Provider-agnostic agent with flexible model routing.

Claude SonnetClaude OpusGPT-5.4
Integrations

Install runtimes and databases on demand

Integrations are set up inside the container when the agent needs them. No pre-configuration required — add PostgreSQL, Playwright, Go, or Rust with a single API call.

PostgreSQL MariaDB Playwright Node.js Python Go Rust Java PHP .NET

Missing an integration for your workflow? Contact us.

Security & Data

Your code stays yours

Credentials are encrypted at rest, containers are sandboxed, and workspace data is ephemeral. Here's how we protect your code and data.

Encrypted credentials

API keys, git tokens, and OAuth credentials are encrypted at rest. Decrypted only when injected into your container at session start.

Environment isolation

Each session runs in an isolated environment with restricted capabilities and a dedicated network. It cannot access internal services.

Ephemeral workspaces

Container data is destroyed after idle timeout. Workspace state (diffs, untracked files) is captured to encrypted storage for session resume, then deleted on cleanup.

No training data

Your code and conversations are never used for AI model training. Data stays in EU infrastructure and is deleted when you delete your account.

FAQ

Frequently asked questions

What happens to my code?
Your code is cloned into an isolated environment when a session starts. The AI agent works directly on the files in that environment. When idle, workspace state (diffs and untracked files) is captured and stored encrypted. The environment is destroyed after the idle timeout. Your original repository is never modified unless the agent pushes changes on your behalf.
Which git providers are supported?
GitHub (including GitHub Enterprise), GitLab (including self-hosted instances), and Bitbucket Cloud. Each provider uses its own personal access token with configurable scopes. You can connect multiple providers simultaneously.
How does session persistence work?
When a container goes idle, the platform captures your file changes, untracked files, and agent conversation history. All data is stored encrypted. When you resume, a fresh container boots, your code is cloned, and the saved state is restored — including the agent's conversation history.
Can I use my own API key?
Yes. You can provide your own Anthropic API key or OpenAI API key. Alternatively, you can connect via your Claude Pro/Max subscription (OAuth) or ChatGPT Plus/Pro subscription (device code flow). Different models may require different credential types — the platform validates this before starting a session.
How do cron jobs work?
Cron jobs run prompts on a schedule (minimum interval: 1 hour). When a cron job fires, it creates a new chat session with the configured prompt, repository, and branch. The session starts a container and the agent executes the task autonomously. You receive push notifications when it completes. If a previous session from the same cron job is still running, the next execution is skipped.
What's in the container?
Python 3.12, Node.js 22, git, curl, ripgrep, jq, and the CLI tools for your git provider (glab, gh, or bkt). The chosen AI agent (Claude Code, Codex, Pi.dev, or OpenCode) is pre-installed. Additional runtimes (Go, Rust, Java, PHP, .NET) and databases (PostgreSQL, MariaDB) can be installed on-demand via integrations.
Is there an API?
Yes. The REST API lets you list repositories, create sessions with prompts, poll session status, fetch messages, and cancel sessions. All requests use Bearer token authentication with gcc_-prefixed API keys. You can also configure webhook endpoints to receive signed event payloads when sessions complete or error. Full OpenAPI docs are available at /api/v1/docs/.
How is pricing structured?
The platform itself is free during beta. You bring your own API key (Anthropic or OpenAI) and pay the model provider directly based on your token usage. The usage analytics dashboard shows per-session cost breakdowns, daily/weekly trends, and 30-day projections so you always know what you're spending.

Read the full documentation →

Ready to let AI work on your codebase?

Free during beta. Bring your own API key. No credit card required.

Create free account → Sign in