agentlaneagentlane
Pages

Every coding agent gets its own running copy of your app

agentlane gives each agent session a lane: its own ports, dev servers and database, and its own iOS simulator, Android emulator or browser, with fast tools to see the app and drive it. Agents use it through the agentlane CLI, an MCP server (look, run, screenshot), and skills and hooks that teach them the loop.

Works with

  • Claude Code
  • Codex
  • Cursor
  • GitHub Copilot
  • Gemini CLI
  • Antigravity

Start in your project

  1. 1Run this in your project

    npx agentlane init

  2. 2Then ask your agent

    Set this project up with agentlane.

Getting started walks through each step.

Claude Code, Codex and Cursor, each shown with its mark, get their own lane from agentlane up: ports, dev server, database and a browser, iOS simulator or Android emulator. Each types an item into its own copy of the Basket app, ticks items off and takes screenshots, then Codex finishes and its lane is released.Claude Code, Codex and Cursor, each shown with its mark, get their own lane from agentlane up: ports, dev server, database and a browser, iOS simulator or Android emulator. Each types an item into its own copy of the Basket app, ticks items off and takes screenshots, then Codex finishes and its lane is released.

Why agentlane

  • Every agent, one install

    Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI and Antigravity get the same CLI, MCP server, skills and hooks from one init.

  • Web, iOS and Android in one tool

    look, run and screenshot work the same on a simulator, an emulator or a browser.

  • The whole environment, per agent

    Ports, dev servers, a database and sign-in, and a device or browser that no other agent touches.

  • Skills that teach the loop

    Six skills take an agent from setting the project up to proving a change with a committed flow or test.

  • It checks the app before your agent trusts it

    doctor confirms that every server runs this worktree's code, and names the fix when one doesn't.

  • Evidence you can open

    Videos, traces, logs and screenshots, plus the Maestro flows and Playwright tests your agent commits.

  • It cleans up

    When a session ends, its lane is released: services stop, databases are torn down, and devices go back to the pool.

  • Local

    No account, no hosted service, and no telemetry of its own.

Fast where agents spend their time

What makes it fast

0.15 s
to read an iOS screen on a warm lane
0.16 s
to read a page in the lane's browser
6.5 s
to get an iOS build from the shared cache, against 91.5 s to build it

The reads are medians of 16 calls and the build is a single run, all on an Apple M4 Max. How these were measured.

The docs, in order

From the first command to the full reference.

  1. 1Start

    1. 1.1Getting startedOne command in your project, then one sentence to your agent.
    2. 1.2Install for your agentWhat init detects, installs and writes for each agent, and the Claude Code plugin.
    3. 1.3Where it runsHosts, platforms and the tools each platform needs.

  2. 2How it works

    1. 2.1What a lane holdsPorts, a device or browser, data and processes, claimed per session and folder.
    2. 2.2How setup decidesWhat the setup skill reads, asks and writes.
    3. 2.3The CLI, the MCP server and the pluginsWhich part an agent uses for what.
    4. 2.4The skillssetup, dev-loop, drive and the verification skills: how an agent learns the project.
    5. 2.5What makes it fastA warm driver, a patched Maestro, a shared build cache and a reused device pool.

  3. 3Features

    1. 3.1Bring a worktree upagentlane up, from a bare worktree to the running app.
    2. 3.2Ports and dev serversEach service gets a free port and a ready check.
    3. 3.3Simulators and emulatorsA pool of simulators and emulators, one per lane.
    4. 3.4Browsers for web appsEach web lane gets its own browser profile and driver session.
    5. 3.5A database per laneGive each lane its own copy of the data, created by up and dropped at release.
    6. 3.6See and drive the applook, run, screenshot and record, on a device or in a browser.
    7. 3.7EvidenceScreenshots, video, traces and logs that outlive the lane.
    8. 3.8DoctorIs this instance worth driving? Read-only checks, each with its fix.
    9. 3.9The build cacheOne build per set of inputs, shared by every worktree.
    10. 3.10The device guard and session hooksHooks that keep agents off each other's devices and browsers.

  4. 4Reference

    1. 4.1CLIEvery command and flag.
    2. 4.2MCP serverThe look, run and screenshot tools over MCP.
    3. 4.3agentlane.jsonThe config file, field by field.
    4. 4.4Environment variablesWhat a lane exports, and the settings you can set.
    5. 4.5TroubleshootingCommon failures and their fixes.
    6. 4.6QuestionsPhysical devices, what leaves your machine, Linux, monorepos, CI and cost.
    7. 4.7Honorable mentionsThe projects agentlane builds on, and the ones it works beside.

Recent releases

Every release

  1. 0.4.1
  2. 0.4.0The patched Maestro build is renamed 2.11.0-agentlane.1, and the toolkit downloads it over https from this repository's release, so first use no longer needs a signed-in gh.