{
  "markdown": "# Agent Ready — Gemini CLI extension / Antigravity plugin\n\nScan any URL for **AI agent-readability** from inside [Gemini CLI](https://geminicli.com) (and,\nafter migration, the Antigravity CLI). It checks sites against the **Vercel Agent Readability\nSpec**, the **llmstxt.org** standard, and agent-protocol manifests (MCP server cards, A2A,\n`agents.json`, `agent-permissions.json`, UCP, x402, NLWeb).\n\nThis extension is a thin wrapper around the existing\n[`agent-ready-mcp`](https://www.npmjs.com/package/agent-ready-mcp) server and the\n[`agent-ready` CLI](https://www.npmjs.com/package/agent-ready-scanner) — both published to npm and\nfetched on demand via `npx`. There is nothing to build.\n\nIt gives Gemini two ways to use Agent Ready:\n\n- **MCP tools** the agent can call autonomously: `scan_site`, `get_scan`, `ask`.\n- **Slash commands** you invoke directly: `/agent-ready:scan`, `:get`, `:list`, `:ask`.\n\n## Requirements\n\n- [Gemini CLI](https://geminicli.com) installed.\n- Node.js ≥ 20 (for the `npx`-launched MCP server and CLI).\n- **No API key required to start** — `scan`/`scan_site` and `ask` work on the free\n  anonymous tier (scan: 3 per 30 days per IP, 25-page depth). A Pro API key from\n  <https://agent-ready.dev/dashboard/api-keys> unlocks `get`/`list` (scan history),\n  deeper 250-page scans, and higher volume.\n\n## Install\n\n```bash\ngemini extensions install https://github.com/mlava/agent-ready-gemini\n```\n\nLocal development (symlink instead of clone):\n\n```bash\ngemini extensions link /path/to/agent-ready-gemini\n```\n\nManage it with the usual commands:\n\n```bash\ngemini extensions list\ngemini extensions disable agent-ready\ngemini extensions uninstall agent-ready\n```\n\n## Configure the API key (optional)\n\nOnly needed for `get`/`list` and deeper scans — `scan` and `ask` work keyless.\n\n```bash\ngemini extensions config agent-ready\n```\n\nThis stores `AGENT_READY_API_KEY` securely (system keychain) and passes it to the MCP server. The\noptional `AGENT_READY_API_URL` overrides the API base URL (defaults to `https://agent-ready.dev`).\n\n> The slash commands shell out to the `agent-ready` CLI, which reads `AGENT_READY_API_KEY` from the\n> environment. If `gemini extensions config` does not propagate the key into shell-command\n> execution in your version, also `export AGENT_READY_API_KEY=ar_live_...` in your shell. The\n> `ask` command never needs a key.\n\n## Usage\n\n### Slash commands\n\n```text\n/agent-ready:scan https://example.com\n/agent-ready:scan https://example.com --page-limit 25\n/agent-ready:get V1StGXR8_Z\n/agent-ready:list\n/agent-ready:list --limit 5\n/agent-ready:ask how is the score calculated?\n/agent-ready:ask what does check S4 do? --type checks\n```\n\nEach command runs the corresponding `agent-ready` CLI subcommand and asks Gemini to summarize the\nJSON output. Gemini will prompt for confirmation before running the shell command.\n\n### Let the agent do it\n\nBecause the MCP server is registered, you can also just ask in natural language:\n\n```text\nScan example.com for AI agent-readability and tell me what to fix first.\n```\n\nGemini will call the `scan_site` tool itself, then `get_scan` if the scan is still running.\n\n## Antigravity CLI\n\nGemini CLI is being replaced by the [Antigravity CLI](https://antigravity.google) (`agy`) for\nindividual-tier users (deadline **June 18, 2026**). Import this extension as a native Antigravity\nplugin with:\n\n```bash\nagy plugin import gemini\n```\n\nThis works cleanly here because the extension stays within the portable subset:\n\n- The MCP server is **local stdio** (`command`/`args`), which carries over unchanged — the\n  `url` → `serverUrl` rename only affects *remote* HTTP MCP servers, and this extension has none.\n- `commands/*.toml` are collapsed into Antigravity **skills** automatically during import.\n- `GEMINI.md` keeps working as the context file.\n- No custom themes or Node-only API dependencies (the only things that don't migrate).\n\n### Using it in Antigravity (after import)\n\nInvocation differs from Gemini CLI — this is expected:\n\n- There is **no `/agent-ready:scan` slash command** in `agy`. The commands become **skills**\n  (named `scan`, `get`, `list`, `ask` under `~/.gemini/config/plugins/agent-ready/skills/`) that the\n  agent invokes, and the **MCP tools** (`scan_site`, `get_scan`, `ask`) are called by the agent\n  directly. Just ask in natural language, e.g. *\"use agent-ready to scan https://example.com\"* or\n  *\"ask agent-ready how the score is calculated\"* (the latter is public, no key).\n- **API key:** the imported `~/.gemini/config/plugins/agent-ready/mcp_config.json` has `env: null`.\n  For keyed `scan`/`get`/`list`, add your key there\n  (`\"env\": { \"AGENT_READY_API_KEY\": \"ar_live_…\" }`) for the MCP tools, and/or\n  `export AGENT_READY_API_KEY=ar_live_…` in your shell for the skill shell-outs. `ask` needs no key.\n\n## What's in here\n\n```\ngemini-extension.json     # manifest: MCP server + settings + contextFileName\nGEMINI.md                 # context auto-loaded into the model\ncommands/agent-ready/     # /agent-ready:scan|get|list|ask (shell out to the CLI)\n```\n\n## Links\n\n- Agent Ready: <https://agent-ready.dev>\n- MCP server: <https://www.npmjs.com/package/agent-ready-mcp>\n- CLI: <https://www.npmjs.com/package/agent-ready-scanner>\n\n## License\n\nMIT © Agent Ready\n",
  "bytes": 5264,
  "sha": "b953a02fcb2925a1152433129f557a382558b5c4645031da41c8f2ee5a04bc79",
  "repo_slug": "mlava/agent-ready-gemini",
  "fonte": "repo",
  "truncated": false,
  "api": "https://agentalog.com/api/listings/plg_mlava_agent_ready_gemini_f1b1b94f/readme"
}