MobileBuildMCP
Gives coding agents MCP tools and a CLI to drive Xcode builds, simulators and tests on iOS and macOS projects.
- Source repo
- getsentry/MobileBuildMCP
- Stars
- ★ 6.5k
- Last updated
- 3d ago
- License
- MIT
- Primary language
- TypeScript
- FA score
- 49/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code
- Cost
- Free, no paid service needed
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- A developer using Cursor, Claude Code or Codex wants the agent to kick off a simulator build and read the resulting errors while editing iOS code.
- Not a fit if
- Teams building iOS/macOS projects on Linux or Windows
- Machines with Xcode below 16 or macOS below 14.5
- Teams that cannot configure code signing yet need device tools
- Source review
- 49/100 · Major gaps
What does this agent do, and when should you use it?
MobileBuildMCP is an MIT-licensed MCP server and CLI published by Sentry for agents working on iOS and macOS projects. It ships as one package in two modes: an MCP server started with mobilebuildmcp mcp that MCP clients connect to, and a terminal CLI that runs the same tools directly. The toolset sits on top of xcodebuild and simulators: simulator build builds a scheme from a project path, --build-for-testing produces portable .xctestproducts, and simulator test runs tests from those products on a named simulator. Stateful operations such as log capture and debugging go through a per-workspace daemon that starts on demand. The runtime boundary is explicit: macOS 14.5+, Xcode 16.x+, and device tools require code signing configured in Xcode.
It wraps Xcode build and test operations as tools an agent can call. Running mobilebuildmcp mcp exposes the MCP server, and mobilebuildmcp tools lists every available tool. A build runs through mobilebuildmcp simulator build --scheme MyApp --project-path ./MyApp.xcodeproj, with --simulator-name picking the simulator; adding --build-for-testing --test-products-path ./MyApp.xctestproducts writes portable test products, which mobilebuildmcp simulator test --test-products-path ... --simulator-name ... later executes. The tool asks xcodebuild to skip macro validation so projects using Swift Macros do not fail to build. mobilebuildmcp init installs two optional agent skills, an MCP Skill and a CLI Skill, that prime the agent to use the tools or navigate the CLI. Long-running stateful work such as log capture and debugging is handled by a per-workspace daemon that auto-starts when needed. Internal runtime error telemetry goes to Sentry, with documented opt-out.
- A developer using Cursor, Claude Code or Codex wants the agent to kick off a simulator build and read the resulting errors while editing iOS code.
- Local verification outside CI: build with
--build-for-testingonce, then rerunsimulator testagainst the stored.xctestproductson a chosen simulator. - A team wants one toolset wired into both an MCP client and shell scripts instead of maintaining two separate build wrappers.
- Maintainers of Swift Macros-heavy projects who keep hitting xcodebuild macro validation errors.
- Teams that want their agent primed with explicit usage instructions by running
mobilebuildmcp initfor the MCP Skill or CLI Skill.
How do you install or deploy this agent?
Pick either method; both give you the CLI and the MCP server.
Homebrew:
brew tap getsentry/xcodebuildmcp
brew install mobilebuildmcpnpm (Node.js 18+ required):
npm install -g mobilebuildmcp@latestVerify the install:
mobilebuildmcp --helpRuntime requirements: macOS 14.5 or later and Xcode 16.x or later; the Homebrew path does not need Node.js. Device tools require code signing to be configured in Xcode first.
How do you use this agent?
Start the MCP server on demand without a global install:
npx -y mobilebuildmcp@latest mcpOr use the installed global binary:
mobilebuildmcp mcpWith the client connected, inspect the tools and run a simulator build:
mobilebuildmcp tools
mobilebuildmcp simulator build --scheme MyApp --project-path ./MyApp.xcodeprojPrepare portable test products first, then run the tests separately:
mobilebuildmcp simulator build --scheme MyApp --project-path ./MyApp.xcodeproj --simulator-name "iPhone 17" --build-for-testing --test-products-path ./MyApp.xctestproducts
mobilebuildmcp simulator test --test-products-path ./MyApp.xctestproducts --simulator-name "iPhone 17"Install the optional agent skills:
mobilebuildmcp initCheck for updates and upgrade in place:
mobilebuildmcp upgrade --check
mobilebuildmcp upgrade --yesWhat are this agent's strengths and limitations?
- One package delivers both a CLI and an MCP server, so agents and shell scripts share the same tool implementation.
--build-for-testingwith--test-products-pathsplits building from running, letting stored.xctestproductsbe reused across test runs.- Deliberately skips xcodebuild macro validation, removing a common failure for projects that use Swift Macros.
- Ships optional MCP Skill and CLI Skill files so the agent is primed on correct usage instead of guessing.
- Installable through Homebrew or npm, and the Homebrew path avoids a Node.js dependency entirely.
- Hard platform requirement: macOS 14.5+ and Xcode 16.x+, so Linux and Windows hosts are excluded.
- Device tools depend on code signing already configured in Xcode; without it those capabilities are unavailable.
- Stateful work depends on a per-workspace daemon that auto-starts, adding a process to account for in deployment and troubleshooting.
- Sends internal runtime error telemetry to Sentry by default; privacy-sensitive teams must follow the documented opt-out.
- Full documentation lives in a separate xcodebuildmcp.com repository, so configuration and troubleshooting details require jumping to another source.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| MobileBuildMCP This agent | 49 · Major gaps | CLIFree | ★ 6.5k | 3d ago | TypeScript | Codex · Claude Code |
| CC Workflow Studio | 50 · Major gaps | IDE extensionFree + model costs | ★ 5.4k | 14d ago | TypeScript | ChatGPT · Codex · Claude Code |
| Microsoft Learn MCP Server | 43 · Major gaps | MCP serverFree | ★ 1.9k | 1mo ago | TypeScript | ChatGPT · Codex · Claude Code · Claude.ai |
| FastAgent | 46 · Major gaps | CLIFree + model costs | ★ 3.9k | today | Python | Codex · OpenAI API · Claude API |
How does FollowAgents rate this agent?
Why each dimension lost points
README shows the tool drives xcodebuild, simulators, devices and a per-workspace daemon, i.e. real side effects on local projects, yet no least-privilege design, per-action confirmation, or rollback guidance is present. Privacy is delegated to an external doc and Sentry telemetry is on by default, limiting data-flow transparency. Dependency security has a concrete positive: a license:check allowlist script and CI validation, hence 2; other items are asserted or absent and are deducted.
package.json states Node>=18, macOS 14.5, Xcode 16 requirements and CI covers build/lint/typecheck/test/schema fixtures, supporting self-consistency; however failure messages and error handling are barely visible in the provided files, so failure_messages is 1.
README clearly separates CLI and MCP server modes and lists Cursor/Claude Code/Codex client setup, fitting the audience; but capability boundaries, trigger conditions and tool-selection logic are not described in-repo and depend on external docs, lowering capability_boundaries and trigger_precision.
Install notes (Homebrew, npm, npx) and example commands are solid, and LICENSE is a full MIT text, hence 3; but there is no CHANGELOG or versioning policy, maintenance responsibility rests on an unverified getsentry brand, known limitations are only scattered (macro validation skip, device signing), and information architecture leans on an external site, so several items are 1-2.
CLI examples (build, test, upgrade) suggest usable outputs and clear marginal value for iOS/macOS agent work; cost-benefit lacks benchmark evidence in-repo (bench scripts exist but no results), so cost_benefit is 1.
Many README claims (AgentAudit Safe badge, CI badge, DeepWiki) point to external resources not verifiable in-repo; fact/inference separation is weak and key behaviors like telemetry and signing are asserted only via links, so all items are 1.
- Sentry runtime telemetry is enabled by default; privacy details live only in an external doc, so verify and consider opting out.
- Tools run xcodebuild, simulator and device operations that can modify local projects and build artifacts; no per-action confirmation or rollback is documented.
- Publisher identity is unverified and safety claims such as the AgentAudit Safe badge cannot be corroborated within the repository.
- No CHANGELOG or versioning policy; the upgrade --yes path should be evaluated carefully.
FAQ
Does it cost anything to use?
Does it send my project data anywhere?
Must an agent go through MCP to use it?
mobilebuildmcp simulator build and similar commands directly; mobilebuildmcp init installs the MCP Skill or CLI Skill to guide it.Why prepare .xctestproducts before testing?
--build-for-testing separates the build from execution, so simulator test can rerun the stored products on a chosen simulator without rebuilding each time.