# ActionSpace AI

ActionSpace helps teams make their products work for AI. It assesses how AI agents discover, understand, connect to, test, and operate a product, then recommends improvements.

Home: https://actionspace.tech/

## MCP

Remote MCP endpoint: https://actionspace.tech/mcp
Transport: Streamable HTTP. The public Safe Scan does not require the beta password or a local server installation.

Add this URL as a remote MCP server in your AI client, initialize the connection, and list its tools.

### Start an assessment

Call `start_agent_readiness_safe_scan` with:

```json
{"url": "https://example.com"}
```

Keep the returned `id` as the assessment ID. This starts a bounded, read-only scan of public pages; it does not submit forms, use credentials, or perform transactions.

### Get the full assessment

Call `get_agent_readiness_assessment` with:

```json
{"assessment_id": "<id returned by the scan>"}
```

If the assessment is queued or running, wait briefly and call again. When `status` is `completed`, the response contains the full report: overall score, grade, five gates (Find, Understand, Connect, Prove, Operate), evidence, recommendations, coverage, and confidence. Report the scan's limitations; something not found in a bounded scan is not proof it does not exist.

The optional `request_agent_readiness_verified_journey` tool records a request for future real-agent verification after a completed scan. It does not execute a journey. Use it only when explicitly requested and the necessary contact details and authorization are provided. Never include passwords, API keys, or other secrets in tool arguments.