Dev tools

Bitrise launches Remote Dev Environments, cloud macOS VMs where Claude Code, Codex and Cursor can build iOS apps

September 17, 2026 at 2:40 AM PT

Bitrise Remote Dev Environments launch image

Image: Bitrise

Why it mattersA coding agent that has to install Xcode before it can build an iOS app burns most of its budget on environment setup, and a pre-configured cloud Mac with the toolchain already cached moves that cost off the meter.

Bitrise launched Remote Dev Environments (RDE) on 16 September through Product Hunt, cloud macOS VMs where AI coding agents run in the same environment as the company's CI. Real Apple hardware in US and EU data centres, Xcode and the Apple toolchain already installed and cached, and named support for Claude Code, Codex, Cursor, VS Code Remote, JetBrains, Xcode, Gemini CLI, Windsurf and any MCP client. An open-source MCP server sits on GitHub at bitrise-io/bitrise-mcp-dev-environments.

What the machines are

Bitrise says the fleet uses M2 Pro, M4 and M4 Pro machines in US and EU data centres and that the stacks are the ones Bitrise CI already runs on, so the environment a developer writes in is the same one the company's CI builds in. Xcode, Simulator, Git, Docker, npm and fastlane are pre-installed. A session is picked from a template, can be archived when work stops and restored later, and the state survives without a meter running. Sessions can be reached over SSH, VS Code Remote, Cursor, JetBrains Remote and VNC for a full macOS desktop.

Where the AI angle sits

Two paths. A developer attaches Claude Code or Codex to a session and runs the agent on their own API keys in the pre-configured environment, or a team routes tickets to agents through the open-source MCP server and spins up parallel headless sessions. Bitrise's page pitches the second at "fleet-scale, not laptop-scale": many agents at once, archived when done, results reviewed at the end.

What Bitrise says it fixes

The company's own framing on the launch page names two problems it says RDE removes. The first is that Bitrise CI reproduction is one machine away rather than "I cannot get this to fail locally": the session is the same stack, so a CI failure is reproduced inside the same VM that produced it. The second is that agents stop burning tokens on environment setup errors, because Xcode, the Simulator and cache are already there and already work. Both are Bitrise's own claims. A team that runs Xcode on a MacBook will have to test whether the same agent finishes a job faster on a rented M4 Pro before it decides on the trade.

The catch worth knowing

The GitHub integration Bitrise names in the docs authenticates git and the GitHub CLI as the developer inside the session, so an agent given the session already has whatever push access that human has. A team that wants an agent to open pull requests without touching main should scope the connected identity narrowly. Bitrise's docs mention this in passing. The open-source MCP server was created in February and last pushed on 16 September, matching the launch, and the repo is public for teams that want to see the surface an agent will call before they commit.

Bitrise has existed as a mobile CI platform for a long time. This is not a new company; it is a mobile-CI vendor betting that agents finishing iOS work on a rented Mac is a bigger business than a person finishing iOS work on their own Mac.

Source

Source: Bitrise

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

More from AI News

MCPJam launches a paid testing and evals platform for MCP servers with Swarms, User Testing and CI/CD

MCPJam went live with a paid platform on 17 September that tests, evaluates and runs CI on MCP servers before they ship, adding Swarms, User Testing, Evals and CI/CD on top of its open-source Inspector, which has 2,207 stars on GitHub.

Source: Product HuntDev tools

CloudX replaces actions/setup-go and cuts its median Go test job from 131 seconds to 41 seconds

CloudX published a drop-in replacement for GitHub's actions/setup-go on 16 September, and says the change cut its median Go test job from 131 seconds to 41 seconds and stopped 86 percent of redundant test package runs.

Source: Hacker NewsDev tools

GitHub rewrote the Copilot agent runtime in 832,378 lines of Rust in about 14 weeks, and one developer led the port

GitHub ported the Copilot agent runtime from TypeScript to Rust in about fourteen and a half weeks, with a single lead developer, 128 landed pull requests, and a Copilot-assisted review loop that generated 832,378 lines of production Rust and 468,689 lines of tests.

Source: Vendor blogDev tools