@extension.dev/mcp

Give your AI agent hands for browser extension development. 30 MCP tools that scaffold, run, inspect, debug, and publish cross-browser extensions.

claude mcp add extension-dev npx @extension.dev/mcp

Works with Claude Code, Claude Desktop, Cursor, and any MCP client.

extension.dev · Extension.js · Templates · Examples · Discord

Why an MCP server for extensions

Extensions fail silently: content scripts that never inject, panels that never open, permissions that return undefined with no error. An agent editing files blind will happily "fix" all of them without noticing none of them work.

These tools give agents eyes on the live browser, so they debug from evidence instead of guessing:

• Scaffold from the 50+ template catalog behind templates.extension.dev, or add a popup, sidebar, or content script to an existing project

• Run the dev server with HMR in Chrome, Edge, Firefox, Brave, Opera, Vivaldi, Yandex, Waterfox, LibreWolf, or any Chromium- or Gecko-based binary, plus Safari on macOS (no HMR yet), no build config

• See the live DOM, unified logs from every extension context, chrome.storage contents, and the loaded-extension list

• Act: evaluate code in any context, trigger the action button and commands, reload the extension, replay events

• Ship: validate the manifest cross-browser, build for production, publish a shareable preview, and promote builds to release channels headlessly

Built on Extension.js, the open-source cross-browser extension framework.

Related MCP server: Chrome Extension MCP Bridge

Clients

Setup

Claude Code

claude mcp add extension-dev npx @extension.dev/mcp

Or install it as a plugin, the MCP server plus the /extension, /extension-add, /extension-debug, and /extension-publish commands in one step:

/plugin marketplace add extensiondev/mcp
/plugin install extension-mcp@extensiondev-mcp

Cursor

Claude Desktop / .mcp.json

{
  "mcpServers": {
    "extension-dev": {
      "command": "npx",
      "args": ["@extension.dev/mcp"]
    }
  }
}

Pair with the skill

This server gives agents hands; @extension.dev/skill gives them judgment: the cross-browser rules, silent-failure gotchas, debugging playbooks, and store checklist, packaged in the open Agent Skills format. With both installed, agents know to verify against the live browser instead of guessing, and these tools make that a one-call operation.

npm i -D @extension.dev/skill
mkdir -p .claude/skills && cp -R node_modules/@extension.dev/skill/skills/extension-dev .claude/skills/

Claude Code project integration

The package ships drop-in instructions, slash commands, and rules for extension projects:

# Rules (how Claude understands your project)
cp node_modules/@extension.dev/mcp/claude/CLAUDE.md ~/my-extension/.claude/CLAUDE.md

# Slash commands (/extension, /extension-add, /extension-debug, /extension-publish)
mkdir -p ~/my-extension/.claude/commands
cp node_modules/@extension.dev/mcp/claude/commands/*.md ~/my-extension/.claude/commands/

Tools

Browser-launching tools (dev, start) shell out to the extension CLI, the project's own node_modules/.bin/extension when present, otherwise npx extension@<pinned> at the version this package is verified against; everything else runs in-process.

Asserting instead of guessing

Every other tool here hands back a reading: a DOM, a log window, an evaluated expression. Turning a reading into "the popup works" was left to the agent, as a string of JavaScript it wrote on the spot, which is the guesswork the paired skill exists to prevent. extension_assert states the expectation and returns the verdict.

{
  "projectPath": "/path/to/extension",
  "expect": [
    { "assert": "background-worker-booted" },
    { "assert": "surface-rendered", "surface": "popup", "selector": "[data-testid=root]" },
    { "assert": "storage-key-present", "key": "settings", "area": "local" },
    { "assert": "console-errors-empty", "context": ["background", "popup"] },
    { "assert": "content-script-injected", "url": "https://shop.example/cart" }
  ]
}

Each check comes back as pass, fail or inconclusive, and the run is a pass only when every check passed. inconclusive is the part that matters: it means this platform cannot cover the question today, and the check carries a settledBy naming the evidence that would answer it. A content script's execution is not observable from outside its isolated world, so content-script-injected passes only on a line the script itself wrote and is inconclusive over a declared match, never a pass. "No console errors" over a session that never built is inconclusive too, because zero errors and zero events are the same number. A read the platform refuses, such as chrome.storage on a session started without allowControl, is inconclusive rather than a failure: nothing was learned about the extension.

The verdict document is the same grammar the preview lane's CI verdict uses (@extension.dev/preview-verdict), with its own contract name and its own check registry, so a document from one lane can never be mistaken for the other's. Each check here names the preview check it is the live-browser counterpart of, and a contract test holds the two grammars together.

Sharing a build in progress

An unpacked extension is unusually hard to hand to someone: the only way to look at a colleague's work-in-progress has been to take their zip and run untrusted code with real browser permissions on your own machine. extension_preview_web with share: true uploads the dist/ it just built and returns a link that renders those exact bytes in the emulator. Whoever opens it installs nothing and signs in to nothing, which is what lets a designer, a PM, or a reviewer into the loop at all. Those bytes run in an isolated sandbox origin or they do not run at all: preview refuses a shared build rather than serving it in its own renderer. Sharing needs auth (extension_auth or EXTENSION_DEV_TOKEN), the link lives 30 days, and DELETEing the returned revokeUrl with the same token kills it early. Re-sharing an unchanged build returns that same link rather than a second one, and only a revoked link is replaced by a different one, because revocation is permanent: the address is burned and never resolves again. That makes revokeUrl the handle to the link you just made, so every share is also appended to .extension.dev/shared-previews.json in the project (gitignored) so it survives losing the tool output. The upload holds up to 2,000 files and about 64MB of text, or roughly 48MB when the build is mostly images, fonts or wasm, which travel base64-encoded. Without share, the tool returns a local-only deep link and uploads nothing.

extension_shares is the other half of that: it lists every link the token has shared, live and dead, with the previewUrl and revokeUrl of each, and revokes one by artifactId or by pasting any of its URLs. Pass projectPath and it reconciles the platform's answer with the project's own record, so a link shared from another machine shows up as remoteOnly and a record with nothing behind it any more shows up under localOnly. It never rewrites the local file.

That is a different job from shipping. Use share for the build you are holding right now; use extension_publish and extension_release_promote below for builds your CI has released.

From preview to store

The platform tools connect agents to extension.dev: extension_auth runs extension.dev's own device flow (you approve the code at extension.dev/device, and GitHub is federated server-side, so no GitHub token ever reaches your machine) and stores a project-scoped token locally (never returned to the agent), extension_publish turns a build your project has already published into a shareable URL, and extension_release_promote promotes a tested build to a release channel from CI or an agent session, no browser required. extension_submit submits a built extension to the Chrome Web Store, Edge Add-ons, and Firefox AMO through extension.dev, which holds your store credentials and dispatches the release from your project's mirror CI, it defaults to a dry run and store credentials are never tool arguments. Safari and the App Store are one paid lane on the platform, so a free workspace is refused there and the other three stores are unaffected. The two verbs are not interchangeable: extension_publish pushes to the extension.dev platform, extension_submit sends the build into a store's review queue, which is irreversible. After a real submission, extension_release_status reads the recorded outcome, per-store credential health, and review state from the project's public registry, so agents and CI can answer "was it approved?" without a console visit. Access tokens live at most 7 days; CI pipelines re-mint them from the console's Access tokens page.

The extension.dev stack

All of it rides on Extension.js, the open-source cross-browser extension framework.

Community

• Join the extension.dev Discord for help and feedback

• Browse production-ready templates at templates.extension.dev

• Follow the platform's public packages on GitHub

• Report Extension.js framework issues on GitHub

License

Apache-2.0 (c) 2026 Cezar Augusto and the extension.dev collaborators. See LICENSE.