Agentastic.dev is the first ADE for Human-Agent collaboration:code isolation, parallel worktrees, Container support, Agent Agnostic, terminal-first, browser-use, code-review, persistent remote SSH, Cloud Agent
Downloador brew install agentasticai/tap/dev
Code, terminals, git, a browser, and 52 AI-agent integrations — together in a single native macOS workspace.
Agentastic.dev is agent-agnostic and ships 52 built-in agent definitions. Run each one in a real terminal, or as a structured chat — 24 of them speak it natively or over ACP.
Each agent works in its own git worktree or container — both Apple containers and Docker are supported. Zero merge conflicts, zero cross-contamination.
A native diff viewer — unified or a continuous side-by-side comparison across every changed file — plus agentic review from Claude, Codex, and CodeRabbit. Catch issues before you merge.
Run agents on any dev server over SSH, or on Modal, Fly, and Vercel — persistent sessions that keep running 24/7, even after your Mac sleeps.
A built-in browser agents can drive — full browser-use, React Grab, and one-click share, right beside your code. Design Mode: click any UI element in a real Chromium window to send its HTML, CSS, and a cropped screenshot straight into your agent's prompt.
A fast native editor, built in — tree-sitter syntax highlighting at a buttery 100 fps. Make the fix yourself without leaving Agentastic.
Use each provider's native CLI, subscription, skills, hooks, and MCP servers. Agentastic auto-discovers installed agents, and any terminal tool can be added as a custom connection.
From idea → code → review.
Start with a prompt or spec that describes what you want to build, change, or fix—or pull it from your issue tracker.
Launch Prime Agent, Claude Code, Codex, Poolside, Gemini, or any of 52 built-in integrations. Agentastic prepares isolated worktrees or containers and keeps agents running 24/7.
Run code review. Review the diff. Everything looks good? Merge and ship—all without leaving Dev.
Free without limits. No sign-up required. No strings attached. We don't charge for the agents. For Agents, you'll need agent subscriptions (Claude Code/Codex/Gemini/Junie/etc) or API access. You pay vendors (Anthropic/OpenAI/Google/JetBrains/etc) directly.
Agentastic.Dev does not collect your source code, file contents, prompts, account details, email, name, or other personally identifiable information for training or analytics. We may send anonymous diagnostic data to Sentry to debug crashes and app hangs, such as an anonymous installation ID, app version, macOS/device context, stack traces, crash or hang details, and network failure metadata.
Each agent runs in its own isolated branch in Git worktree or docker container; they can work simultaneously without conflicts. You can monitor all agents in real-time and switch between them instantly.
Yes, each Agentastic workspace is a new git worktree. This allows each coding agent to work in complete isolation without conflicts.
We ship 52 built-in agent definitions, including Prime Agent, Claude Code, Codex, Poolside, MiMo, Qoder, OpenClaude, Ante, Oh My Pi, OpenClaw, Gemini, Hermes Agent, Cursor, Copilot, OpenCode, CodeRabbit, Greptile, and more. Any terminal-based agent can also be added as a custom connection. Want to see something else? Email us at hi@agentastic.ai.
Agentastic.dev requires macOS 14.0 (Sonoma) or later and runs on both Apple Silicon and Intel processors.
Right now Dev is still in closed development. But our goal is to build an open, community-driven ecosystem, and extensibility is on the roadmap. If you'd like to shape what that looks like, ping us on Twitter or join our Discord!
Dev supports hooks, including .agentastic/setup.sh, which executes upon worktree creation. You can setup your environment variables or copy them using this hook. See the documentation for more examples.
