MCP tool reference
AI Retrievability by Norg — production connector
Every tool the live AI Retrievability by Norg MCP connector exposes, what each one reads or changes, and what data the connector collects. For how sign-in works, see the authentication reference; for what the product does end to end, see the AI Retrievability by Norg guide.
The connector
One HTTPS endpoint, reachable by adding the URL alone — no pre-shared credential required.
URL: https://content-craft-api.norg.ai/v2/mcp/express
Transport: Streamable HTTP, stateless — no session state is kept between calls.
Auth: OAuth 2.1 with mandatory PKCE, scope mcp:onboarding. The server supports Dynamic Client Registration (RFC 7591), Client ID Metadata Documents, and a pre-registered confidential client — full mechanics, discovery probes and troubleshooting are in the enterprise authentication reference.
curl -s https://content-craft-api.norg.ai/v2/mcp/express \
-X POST -H "Content-Type: application/json" \
-d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{}}'
# -> 401, with a WWW-Authenticate header pointing at the discovery documentTools
Nineteen tools, and no others. Eight are read-only; six write without deleting or replacing anything; five are destructive (delete, replace, or remove something already live) and Claude confirms with the customer before using one.
| Tool | Access | What it does |
|---|---|---|
Get Plan Details get_onboarding_manifest | read-only | Shows the plan, its price, and the first decision: NORG-hosted or self-hosted. |
Set Business Details set_business_profile | write | Saves the business name and homepage URL to mirror. |
Start Free Trial start_payment | write | Gets a secure checkout link to start the free trial. |
Confirm Payment confirm_payment | write | Confirms the trial using the code shown after checkout, and begins provisioning the mirror. |
Check Setup Status get_onboarding_status | read-only | Reports how far setup has got, and what to do next. |
Change Mirror Hosting switch_hosting | destructive | Moves the mirror between NORG hosting and the customer’s own CDN. |
Install On Your CDN install_on_cdn | destructive | Deploys the mirror worker into the customer’s own CDN account, using an API credential they supply. |
Remove NORG From Your CDN uninstall_from_cdn | destructive | Removes the worker and its routes from the customer’s CDN account; the subscription is unaffected. |
Finish CDN Install cdn_install_status | write | Completes a CloudFront or Bunny install once the stack has built or the CNAME resolves. Safe to call repeatedly. |
View Mirror Homepage get_mirror_content | read-only | Shows the mirror homepage in three parts: the crawled text, the research text NORG wrote, and the Q&A. |
Edit Mirror Homepage update_mirror_content | destructive | Replaces the research text and/or Q&A on the mirror homepage and republishes it. The crawled text is read-only. |
View My Account my_account | read-only | Shows the plan, price, currency and how many trial days remain. |
Cancel Subscription cancel_membership | destructive | Ends the subscription at the end of the current trial or billing period. |
View Mirrored Pages get_mirrored_pages | read-only | Lists every page currently being mirrored. |
View AI Crawler Analytics get_crawler_analytics | read-only | Shows AI crawler visit totals for the last 30 days, grouped by crawler company. |
Recrawl My Mirror trigger_recrawl | write | Re-crawls the site and refreshes the mirror now. Rate-limited to once every 7 days. |
View Queue get_job_queue | read-only | Reports how busy the setup and crawl queues are, in aggregate. |
Run AI Search Audit run_ai_search_audit | write | Starts a free audit of how discoverable the site currently is to AI search. |
Check AI Search Audit get_ai_search_audit | read-only | Checks the results of an AI search audit that has been run. |
Example prompts
Nothing here needs exact phrasing — Claude maps a natural-language request to the right tool call. These are representative, not a fixed command syntax.
| What you could ask | Tool(s) it reaches |
|---|---|
| “What does the AI Retrievability plan include, and what does it cost?” | get_onboarding_manifest |
| “Set up AI Retrievability for Acme Coffee at https://example.com” | set_business_profile |
| “Start my free trial for that site, then tell me when the mirror is live.” | start_payment, get_onboarding_status |
| “Move my mirror onto my own Cloudflare account.” | switch_hosting, get_onboarding_status |
| “Change the mirror title to Acme Coffee review test, keep the body the same.” | update_mirror_content |
| “Add a Q&A entry answering 'Do you ship internationally?'” | update_mirror_content |
| “List all the pages you are mirroring for me.” | get_mirrored_pages |
| “How much AI crawler traffic has my site gotten in the last 30 days?” | get_crawler_analytics |
| “Refresh my mirror now, my site has changed.” | trigger_recrawl |
| “Run a free AI search audit on my site.” | run_ai_search_audit |
Data handling
- Account identity from the OAuth token (subject, email).
- The business name and public homepage URL the customer provides.
- Stripe customer/subscription identifiers for trial and billing status — never card data.
- A CDN API credential, only if the customer chooses to self-host — encrypted at rest, used solely to install, update, or remove the edge worker, and never logged or echoed back.
- Card details — entered only on Stripe’s own hosted checkout page.
- Conversation content.
- A submitted URL is validated before it is ever fetched: non-HTTP(S) schemes and loopback, private, link-local, and cloud-metadata addresses are rejected (SSRF protection).
- Payment confirmation is verified server-side against a Stripe webhook and bound to the paying account; a confirmation code cannot be replayed or transferred to a different account.
- The mirror is served only to AI agents whose source IP is verified against that vendor’s published address ranges. A spoofed user-agent from any other IP gets the customer’s ordinary site, unchanged — anti-cloaking by construction, not by policy.
Talk to us
For a question about a specific tool, or to report a page that no longer matches the connector, write to solutions@norg.ai and include:
- the tool name and the exact request that produced the result in question
- the workspace or account it happened on
- the time it happened, if it is a live-behaviour question
For an administrator approving the connector, see the enterprise authentication reference. For the per-CDN technical detail behind how NORG serves content, see the edge routing white paper.