Use the Register from an AI assistant or agent
The Register runs a read-only server that AI assistants and agents can query directly, over the Model Context Protocol (the open standard that Claude, ChatGPT, Cursor, VS Code and others use to connect to outside data). Ask "have coding agents deleted production data, and which controls were missing?" and the answer comes back with case IDs, typed claims and links to the primary records.
Connect
Server address:
https://secureautonomous.com/api/mcp
No account, key or sign-in. In an assistant that supports remote servers (for example Claude's custom connectors, ChatGPT's connectors, Cursor, or VS Code), add a new server and paste the address. For tools configured by file, the entry is:
{
"mcpServers": {
"secure-autonomous-register": { "type": "http", "url": "https://secureautonomous.com/api/mcp" }
}
}
What it can do
| Tool | What it returns |
|---|---|
search_cases | Cases matching keywords, a world (physical, bridge, software, transaction, boundary), or a control that was missing or failed |
get_case | A whole case: every claim with its type and source links, outcome, controls, the Register's labeled interpretation, right-of-reply status, and a citation |
list_controls | The 28 controls, with how many cases record each as failed, absent or held |
get_control | One control, its cases, and the regulations and standards that require, imply or recommend it |
find_rules | Regulations and standards by keyword, jurisdiction or control |
recent_signals | The nightly stream of new primary records: vulnerability entries, recalls, rules |
Ground rules the server gives every assistant
- Counts describe the Register's curated cases, not autonomous systems in general. They are never rates or estimates.
- An assertion is reported as the named party's claim, not as fact. The Register's interpretation is labeled as such.
- Cite as: Secure Autonomous Register, case ID, with the case's link.
- The server shows exactly what this site shows. Cases held for right of reply are not served until they are published.
Other ways in
For crawlers and retrieval systems: llms.txt and llms-full.txt. For analysis: the full export in CSV and JSON, including controls, rules and signals, each case's own case.json, and the signals RSS feed. Everything is published under Creative Commons Attribution 4.0.