Got AI Policy now connects to your AI assistant

Here is a simple question that should be easy to answer, but usually is not: does Toronto have a published AI policy? What about every capital city in Canada, or every municipality in Alberta that is still "in progress"?

If you are writing a council memo, reviewing a vendor clause, or prepping a governance brief, you should be able to ask those questions where you are already working and get a sourced answer back. Not after opening six tabs, not after hunting through old PDFs, and not after guessing which search result is still current.

That is the idea behind the MCP integration at https://gotaipolicy.ca/mcp — and since launch we have expanded the toolset so your assistant can now roll up, drill down, and filter the registry the same way you can on the site.

What is MCP?

MCP stands for Model Context Protocol. Think of it as a universal adapter that lets an AI assistant talk to a live data source the same way a browser talks to a website.

Instead of copying and pasting static text into a chat window, an MCP client can query a server for real-time, structured information. The server returns facts with links, and the assistant can cite those facts while you work.

For Got AI Policy, the MCP server exposes the same public registry data you see on the site: municipality names, policy status, province, policy type, provincial and territorial policies, and the source URL where a human verified the record. It does not expose member content, drafts, forum posts, or account data. It is read-only and anonymous by design.

What's new: nine tools, not one

At launch the server shipped with a small handful of lookup tools. It now advertises nine, so an assistant can move from a national snapshot all the way down to a single municipality's source document without you having to switch contexts:

• get_registry_snapshot — national totals: how many municipalities have a published policy, are in progress, or have no policy on record.
• list_provinces — province-by-province coverage counts for all 10 provinces and 3 territories.
• list_featured_cities — the curated shortlist of Canada's 25 largest cities plus every provincial and territorial capital.
• search_municipalities — find cities by name, province, or status.
• search_by_policy_type — discover municipalities by policy type (standalone directive, embedded clause, procurement guardrail, governance framework, and more).
• list_municipalities_by_status — complete filtered lists for a given status, e.g. every "published" municipality in Ontario.
• get_municipality — a single city's full record: policy type, source URLs, last AI-scouted date, human-verification status, and secondary evidence.
• get_provincial_policy — provincial and territorial AI-use policies that sit above the municipal layer.
• compare_municipalities — side-by-side comparison of two or more municipalities.

Every tool returns the source URL alongside the answer, so your assistant can cite the underlying document rather than paraphrasing it.

What you can ask

Once connected, your assistant can answer questions like:

• How many Canadian municipalities have a published AI policy right now?
• Which municipalities in Ontario have a published AI policy, and which are still in progress?
• List every capital city with a standalone AI directive.
• Which Alberta municipalities have no public AI policy found?
• Does Vancouver have a standalone AI policy or an embedded clause? What is the source URL?
• Compare the AI policy status of Ottawa, Calgary, and Winnipeg.
• What does British Columbia's provincial AI-use policy say about generative AI?
• Show me every municipality using a procurement-focused AI policy.

Every answer is tied to the live registry. If the status changes, the next query reflects the updated record.

How to connect

The fastest path is to visit https://gotaipolicy.ca/mcp for per-client instructions. The short version is:

• Copy the endpoint: https://gotaipolicy.ca/mcp/server
• Open your MCP-compatible client (Claude, ChatGPT, Cursor, Notion, or any other client that supports MCP).
• Add a new MCP server and paste the endpoint.
• Leave authentication blank if the client allows it. Some clients, such as Notion, require you to sign in to every MCP server before adding it — if prompted, you can use your normal Got AI Policy email and password. The server still treats every request as anonymous public access.

No API key is required. No paid tier is required for the MCP server. The registry data is public, and the server is public.

Why we built this

Got AI Policy was never meant to be just a website you visit. It is meant to be a reference layer that shows up wherever AI governance work happens.

Right now, that work happens inside memos, RFPs, council reports, vendor calls, Slack threads, policy drafts, and comparison spreadsheets. Every time a professional has to leave that context to look up a source, friction wins and good questions go unanswered.

MCP removes a small but meaningful slice of that friction. It lets practitioners keep their momentum while still pulling from a verified, dated source of truth.

The longer view: into the work layer

This launch is one of several integrations we are pursuing because we believe AI governance should live at the layer where people actually work.

A standalone website is useful for discovery. But a standalone website is not enough when the questions are continuous: Is this policy still current? Did that municipality update its guidance? What did peers do after that incident? What should our procurement clause say? How does our draft compare?

"AI governance is not a document you write once. It is continuous work that evolves as the technology, the edge cases, and the questions change."

MCP is one way we are trying to make that continuous work easier. It is not the only way, and it is not the final way. It is an early bridge between the public registry and the private workspace.

What this means for practitioners and organizations

If you are a municipal policy writer, a consultant, a procurement officer, a vendor, or a journalist, the MCP server means you can treat Got AI Policy as an embedded reference rather than a destination.

If you are an organization, it means your team can query Canadian municipal AI governance evidence without everyone building the same manual research habit from scratch.

And if you are building a system, a workflow, or a product that touches AI governance, this is the shape of the interface we want to offer: open, source-first, and scoped to the public data that everyone should be able to rely on.

The next steps

If you want to try it, start at https://gotaipolicy.ca/mcp. The setup guide covers Claude, ChatGPT, Cursor, Notion, and other MCP clients.

If you are not on a paid tier yet and want to do deeper work with the registry — comparing municipalities, exporting policy reviews, or tracking changes — the Starter plan is built for solo practitioners who need more than the free tracker limits. You can see the details at https://gotaipolicy.ca/pricing.

Frequently asked questions

What is MCP?

MCP stands for Model Context Protocol. It is an open standard that lets an AI assistant query a live data source the same way a browser requests a web page. For Got AI Policy, it means your assistant can ask the registry for current municipality records and get back verified facts with source links.

What tools does the server expose?

Nine, as of this update: get_registry_snapshot, list_provinces, list_featured_cities, search_municipalities, search_by_policy_type, list_municipalities_by_status, get_municipality, get_provincial_policy, and compare_municipalities. Most MCP clients auto-discover the list — you do not have to configure them individually.

How current is the data the server returns?

The server queries the same live registry that powers the website. Automated scouts re-check known source URLs weekly, and human reviewers verify newly-submitted policies. Each municipality response includes the last AI-scouted date and, where applicable, the last human-verification date, so your assistant can judge freshness before citing an answer.

Do I need a Got AI Policy account to use the MCP server?

No. The MCP server is public and anonymous by design. It exposes the same registry data anyone can browse at https://gotaipolicy.ca/municipalities. Some clients, such as Notion, require you to sign in to every MCP server before adding it — if prompted, you can use your normal Got AI Policy email and password. That sign-in is a client requirement, not a server requirement, and it does not unlock any private data.

Is the MCP server free?

Yes. The MCP server is free to use. There is no API key, no paid tier, and no rate limit for normal individual use. It queries public registry data, so the cost of running it is covered as part of keeping the registry open.

What data can my assistant access?

Only the public registry data: municipality names, provinces, AI policy status, policy type, source URLs, verification dates, provincial and territorial AI-use policies, and the featured-cities shortlist. The MCP server cannot access member library content, uploaded policy drafts, private reviews, forum posts, direct messages, or user account data. It is read-only and cannot change anything in the registry.

Can the server write, submit, or edit policies?

No. Every advertised tool is read-only. Submissions still go through the human-reviewed evidence form on the website so a person verifies the source before it enters the registry.

Is it safe to connect an AI assistant to the registry?

Yes, because the server is read-only and scoped. Every request is treated as an anonymous visitor and filtered by the same row-level security rules that protect the website. The server cannot write data, and it will not return anything that is not already public. As with any AI assistant, you should still verify important answers against the source URL returned by each tool.

Which AI assistants or clients work with the MCP server?

Any client that supports the Streamable HTTP MCP transport can connect. The setup guide at https://gotaipolicy.ca/mcp includes instructions for Claude, ChatGPT, Cursor, Notion, VS Code, Zed, and other clients. If your client only supports stdio servers, you can wrap the endpoint with a proxy such as mcp-remote.

Why does Notion ask me to sign in when other clients do not?

Notion requires authentication for every MCP server, even public ones. That is a Notion policy, not a Got AI Policy policy. When Notion prompts you, sign in with your normal Got AI Policy email and password, or create a free account first. The server continues to treat every request as anonymous public access, so signing in does not change what the assistant can see.

Does the server cover provincial and territorial policies too?

Yes. The get_provincial_policy tool returns published AI-use policies for the 10 provinces and 3 territories that sit above the municipal layer — useful when a municipal question depends on higher-order guidance from Ontario's OCIO, Quebec's Law 25 directives, or BC's OIPC guidance.

How should I cite an answer that came through MCP?

Cite the source URL returned by the tool, not the assistant. Every municipality and provincial policy response includes a direct link to the underlying document — treat that as the citation of record. The registry entry is the pointer; the source document is the truth.

Resources and further reading