Gateway Catalyst makes your licensed research reachable from the AI tools your subscribers work in. Their licence is checked on every request, your name is on the answer, and you can still see it happened.
275000+ users
on Content Catalyst platforms
20+ years
building for analyst research
If a client has already wired up its own connector, that’s only one of the routes your research travels, and usually it was built without anyone asking you. Routing through you gives them a better answer than the one they built, which is why they’d move.
Someone at a client firm links the assistant they already use to your library, signing in with the portal credentials they have. It usually involves their IT team.
Questions in plain language, answered from your research, limited to what their licence covers, and citing the reports behind the answer.
No new tool, no second login. Where possible the citation links straight back to the report or dataset on your platform, so the original is one click away.
Start with the client who asked. Connections can be revoked one subscriber at a time, so a pilot can end cleanly if it doesn’t work out.
Three views of the same connection.
What your subscriber sees.
Their licence is checked before anything is retrieved.
A question asked in the assistant they already use, answered from your research, with the source named and, where possible, linked back.
Your dashboard.
The same numbers an account manager takes into a renewal.
Browses, reads, citations and searches, with a weekly trend and every product ranked by engagement.
Interrogate your data.
Go past the dashboard's fixed views and ask your own questions.
Your own assistant, pointed at your usage. Top searches, and the dead ends worth reading as content gaps.
The usual worry about research reaching a client through their own AI is that you go blind the moment it does. You don’t.
Not "your content was cited" as a feature, but a number, per title, that an account manager can take into a renewal conversation.
Sessions, searches, what was browsed, what was read in full, and which products earned citations. Sortable and downloadable.
A broad search returning an undifferentiated wall of results; a subject your subscribers keep reaching for that you haven't covered. Commissioning material as much as account material.
Your administrators can point their own AI assistant at your usage data and ask it things, rather than reading a fixed report. Session summaries, ranked gaps, whatever they put to it.
Your library is queried in place, never exported or synced, and never used to train a model. Only the passages that answer a licensed subscriber’s question leave it. Your existing firewall, session and access-logging controls still apply.
On or off across the whole site. On or off per tool. Controlled at site, user and admin-role level. And an individual subscriber’s connection can be revoked without touching anyone else’s.
Enforcement is per user and per request, at category and report level. Unlicensed content returns nothing: no summary, no citation, no partial answer. Access ends with the contract, and licence changes apply from the next request.
Private reports are discoverable only under licence. Embargoed and unpublished material is never exposed. Publication dates travel with your content, so recent analysis comes first without older reports being pushed out just for being old.
A generated answer isn’t guaranteed correct. Each one cites the reports behind it, and the citations are counted.
Is this MCP?
Yes. Gateway Catalyst is built on the Model Context Protocol, the open standard that lets AI tools connect to external content, so it works with any assistant that supports it rather than a list we approve. Being an open standard is the point: MCP is not the only one of its kind and will not be the last, and your library stays where it is either way.
Couldn't our developers just build a connector?
They could, and the connector is the straightforward part. The work is what surrounds it: keeping entitlements true across many clients on different terms, and keeping usage visible once research is read somewhere you do not own.
Our clients already have their own connectors pulling in research, including ours. Why route through us?
What reaches them that way arrives unlicensed, uncredited and invisible to you. Routed through you, it arrives with your name on it and a record that it happened.
Which AI tools does it work with?
Any that support MCP. Your subscribers keep whichever assistant their firm has standardised on, and you are not tied to a provider either.
Does this pull subscribers away from our portal?
Where possible, citations link back to the report or dataset itself, so an answer is a route into your platform rather than a substitute for it. A subscriber who wants the underlying analysis still has to come and read it.
Does our content leave our platform?
Your library stays where it is: queried in place, never exported or synced, and never used for training. What leaves is the passage that answers a licensed subscriber’s question.
Can we switch it off?
Yes, across the site, per tool, or for one subscriber.
See both sides of it
A subscriber asking a question in their own AI, and the usage that comes back to you.
Working out where to start? Read our thinking on licensed research and AI answers first.