If you already work in Claude, Cursor, or another MCP-compatible client, you can query your organisation's own knowledge base from there: ask what a protocol says, get the passage back, and see which questions your teams get wrong most often.
Ask a model what your escalation policy says and it answers from general training data, because it has never read your escalation policy. Connect it to PubDaily over MCP and the same question is answered from the documents you uploaded, with the passage attached.
Request early access below and we issue a key scoped to your workspace.
Point your MCP-compatible client (Claude, Cursor, or your own agent) at the PubDaily MCP endpoint with your key.
Every response comes back cited and confidence-scored, the same schema documented on the API reference.
MCP is a thin, standard interface over the same endpoint documented in the API reference, not a separate product or a second knowledge base.
Every answer traces back to a document your organisation uploaded, and names the passage it came from.
Ask which folders a team is weakest on, or which question your organisation misses most, without opening the dashboard.
A key only ever reaches your own knowledge base and your own results, under the same access rules as the dashboard.
You should be comfortable adding an MCP server to a client like Claude Desktop or Cursor. No app to build if you're just asking questions yourself.
Yes. MCP is an interface on top of the same endpoint documented in the API reference, not a separate product or a separate knowledge base.
Only the documents in your knowledge base and the results your key is scoped to. It has no access to anything outside your workspace.
It's early access while we work with the first cohort of users. Request access and we'll reach out to get you set up.
We're onboarding a first cohort directly. Tell us what client you're using and what you'd want to ask.
Request early access →