Agent Ready

See your site the way search engines, AI answers and AI agents see it, and fix what stops them.

Agent Ready is a free, open-source Chrome extension, with a local MCP server and a set of skills for coding agents. It checks any page for three things:

• SEO: can search engines crawl, index and understand the page?
• GEO: can AI answer engines such as ChatGPT, Gemini and Perplexity reach, read and cite it?
• AXO: can an AI agent use the site? Can it name the controls, call the forms as tools and finish a task like finding a price?

It needs no account, sends nothing to us and has no telemetry. The checks come from the audits WorkPods runs for its clients, with the same ids, severities and wording.

_{Real results for allbirds.com, used as an example only. More in Using the extension.}

What it does

Free, with no keys:

• Checks the page you are on. Each check passes, fails, partly passes or is untested, and a failure carries a severity. Show on page marks the element, and Copy fix copies a prompt for your coding agent.
• See as AI puts the raw HTML that AI crawlers read next to the page people see: word counts, schema types and text. If your copy only appears after JavaScript runs, it shows here.
• Reads robots.txt, sitemaps, llms.txt and agent manifests, and says which AI crawlers may fetch the site.
• Crawls up to 200 pages, one at a time and within robots.txt, to find broken links, duplicates, orphan pages, soft 404s and hreflang errors.
• Compares your page with a competitor's, check by check.
• Reads an AI answer open in your tab (ChatGPT, Gemini, Claude, Perplexity, Google AI Mode or an AI Overview) and marks which brands it names and which sites it cites.
• Runs a panel of buyer prompts in Assist mode: the extension types each prompt, you press Enter, and it scores the answer.
• Sends an AI agent through a task on your site, such as finding a price. The agent stops before a form is sent, an order is placed or anyone signs in.

With your own keys, kept encrypted on your device under a monthly cap you set:

• DataForSEO: keyword ideas with volume and difficulty, competitor discovery, keyword gap, backlinks and link gap.
• Cloro: prompt panels that run unattended on ChatGPT, Gemini and Google AI Mode.
• Anthropic, OpenAI or Gemini: the model that runs agent journeys from the panel. Your coding agent can run journeys instead, through the bridge.

From your coding agent: Claude Code, Codex, Cursor and other MCP clients can run every check through the local bridge. The skills turn the results into an audit report, and then into code changes in your repository.

Get started

Install Agent Ready from the Chrome Web Store.

The npm bridge package provides agent and CLI commands. See Agents and skills for setup. TESTING.md covers unpacked builds for development and release testing.

1. Install the extension, open its side panel and click Check this page: Getting started.
2. Learn what each tab does: Using the extension.
3. Connect your coding agent and add the skills: Agents and skills.

# Give Claude Code the Agent Ready tools, then turn on Agent connection in the extension's settings
claude mcp add agent-ready -- npx -y @workpods/agent-ready mcp

# Check a page from a terminal or a CI job, with no browser
npx @workpods/agent-ready check https://example.com --fail-on high

Documentation

Guide	Read it to
Getting started	Install the extension and run your first check
Using the extension	Work through each tab, from This page to Journeys
How the checks work	Read a score, a status and a severity, and know what "untested" means
Keys and spending	Set up the vault, add DataForSEO, Cloro or a model key, and cap what you pay
Agents and skills	Connect Claude Code, Codex or Cursor, and use the seven skills
Command line and CI	Check pages from a terminal and fail a build when a page regresses
Privacy and permissions	See what is stored, what leaves your device and why each permission is asked

How it is built

packages/core/    The checks, in plain TypeScript. They take a parsed page, headers and files and
                  never call Chrome, so the extension, the bridge and the CLI share them.
extension/        The Chrome extension (Manifest V3, WXT and React): side panel, service worker,
                  content scripts and the full-page crawl report.
bridge/           The npm package @workpods/agent-ready: the MCP server and the CLI.
plugin/           The skills for Claude Code and Codex.
vendor/workpods/  The check registry, theme and fonts, copied from WorkPods.

To work on it you need Node 22 or later:

npm install
npm run build -w extension   # then Load unpacked: extension/.output/chrome-mv3
npm run check                # format, typecheck, tests, skills lint and build

Read AGENTS.md before you change anything; it has the rules every change follows. Two matter most. The extension never calls a WorkPods server. API keys are decrypted only in the service worker and never reach a page, the bridge, a log or an export. Tests replay recorded provider responses and never call a paid API.

Found a bug or a verdict you think is wrong? Open an issue with the page URL, what you expected and a screenshot of the panel. Never paste an API key into an issue.

Firefox builds

Firefox desktop 140 or later is supported by the Firefox package. Build it with npm run zip:firefox -w extension. Chrome builds use a side panel and service worker; Firefox uses its sidebar and a background event page. Firefox stores private records in extension-origin IndexedDB and asks for data-consent permissions at install time. Optional cloud providers still require your own keys. Chrome's built-in model is unavailable in Firefox.

The local MCP bridge needs the exact Firefox installation origin from Settings; see Firefox bridge setup.

Licence

Apache-2.0, copyright WorkPods. Written by Shefali Munjal. The WorkPods name and logo are trademarks and are not covered by the licence; see NOTICE.

If you use Agent Ready in a commercial product or service, please credit "Agent Ready by WorkPods" publicly, with a link to this repository. The licence does not require it, but it is how we can keep the tool free.