Safe autonomous execution
Each agent runs in its own isolated container with an egress firewall, so you can confidently disable protections like confirmation prompts and let agents work autonomously without risking your host environment.
Safe autonomous execution
Each agent runs in its own isolated container with an egress firewall, so you can confidently disable protections like confirmation prompts and let agents work autonomously without risking your host environment.
Isolated databases and services
Every instance gets its own database, cache, and backing services via Docker Compose — no cross-talk between agents, no collision with your local dev setup.
Browser testing built in
Per-instance Playwright and headless Chromium (CDP) sidecars give each agent its own browser session for E2E testing, visual verification, and browser automation via MCP.
Parallel by default
Spin up as many instances as your machine can handle. Run tests, builds, and agents in parallel across fully independent environments with deterministic port mappings.
Live previews
Expose dev server ports from any instance and preview running applications directly from your host — each container gets its own mapped ports.
Orchestrate an agent fleet
Coordinate any amount of work across an agent fleet. The coordinator resolves dependencies, sequences merges, and runs watchdog checks — from a handful of issues to an entire backlog.
Multi-instance management
Run multiple isolated devcontainers in parallel, each with their own workspace and state. Instances auto-assign planet names with deterministic port mappings.
Claude Code integration
Each instance has Claude Code pre-installed with configurable model, effort level, hooks for progress logging, and transcript export.
Network firewall
iptables-based allowlist restricts egress to GitHub, npm, Anthropic, and your project-specific domains.
Autonomous agents
Run cspace up <name> --prompt-file <path> to fully autonomously resolve a GitHub issue: explore code, design architecture, implement, verify, and ship a PR.
Multi-agent coordination
The coordinator prompt manages parallel agents across issues with dependency tracking and merge ordering.