Give every agent a real machine.
The foundational building block for your Software Factory. Isolated macOS and Linux machines with your repo, tools and devices already in place. Orchestrate them from any agent, and step in when it matters.
Sandboxes run code. Agents ship apps.
An agent needs what your engineers need: the real OS, the real toolchain, and a device to run the app on.
Agents that can see the app.
Real macOS on Apple Silicon with Xcode, and Linux with Android tooling. Agents work against a live simulator or emulator, not a guess.
Device sessions. Boot a simulator or emulator with the session and watch it in the browser.
Eyes and hands. Agents screenshot, click, type, scroll and drag through MCP or the API.
Turn a build into a link.NewReviewers open it on a live device, no Bitrise account needed.
Everything your agents need. Nothing they wait for.
Fan out agents
One MCP call per task. One isolated VM per agent.
Warm pools
AlphaBooted and warmed up before anyone asks.
Devices in the loop
Agents see and tap the app. Reviewers get a link.
Know when it's your turn
Live status, then a browser notification.
Spend on purpose
Sessions shut themselves down. Restore later, or delete for good.
Lock down what agents reach. Speed up how they start.
What we're building next for teams that run agents at scale.
Network filtering
Default-deny egress for every session with deep packet inspection (DPI). Allow only the hosts, ports and requests an agent needs.
Secrets the agent never sees
Tokens are added on the host, outside the VM, so a prompt injection has nothing to leak.
Custom VM images
Bake your dependencies into your own image, and start every session and warm pool from it.
Step in when it matters.
Every agent shows Working, Needs input or Idle. Answer in the browser terminal, or open the session in VS Code, VNC or SSH.
Build your agent platform on it.
Everything the UI does is on the API. Orchestrate from any MCP client, the CLI or your own scripts.
Any MCP client becomes an orchestrator
Open source (MIT) and hosted at mcp-rde.bitrise.io. Sign in with Bitrise, no token needed.
One command to a cloud Claude Code
Run bitrise-cli rde claude in any repository. It clones your branch into a fresh session and opens Claude Code there.
Exit anytime. --resume picks it back up.
Build your own agent platform Experimental
Call api.bitrise.io/rde/v1 with a personal access token. Create sessions, run commands and mint preview links from your own services.
A bigger machine, right in VS Code
Pick a repo, branch and machine. The extension opens a remote VS Code window on the session, with your git identity and extensions.
RDE is the inner loop of your software factory.
The agent codes, builds, tests and debugs on its own machine. When it's ready, the work hands off to Bitrise CI and on to your users.
Three commands to Claude Code on a cloud machine.
Run them from any git repository. RDE clones your branch and drops you straight into Claude Code.
Pay for minutes, up to a monthly cap.
Included minutes come first, then pay as you go up to a monthly cap. After the cap, usage is free for the rest of the month.
FAQ
Which agents does it work with?
Claude Code is built in, with live status and notifications. Any MCP client, like Cursor, Antigravity CLI, VS Code, Windsurf or Copilot, can create and drive sessions.
How fast does a session start?
Templates run your warmup script once. Warm pools (alpha) keep machines booted, so a new session is handed over instead of starting cold.
Can agents test mobile apps?
Yes. Sessions boot an iOS simulator or Android emulator, and agents can screenshot, click and type on it.
How does pricing work?
Included minutes first, then pay per minute up to a monthly cap. After the cap, the rest of the month is free.
Give your agents somewhere real to work.
Start with one command, or connect the agent you already use.