Connect your own feeds and APIs, drop any value into a prompt.
You maintain RSS feeds, API endpoints and data blocks yourself: create them, test them live, pick the fields you need. A block such as [chart_12m] fetches the real data for each name at writing time.
What you can do with it
- Add RSS feeds under “Research & Inspiration”: the URL is checked first, for entries and for duplicates. New feeds join the next automatic scan.
- Import many feeds at once, one title and URL per line.
- Define API endpoints: method, parameters as single value, CSV or array, date expressions such as today-30d, custom headers, path parameters and the path to the result list.
- Refer to API keys in parameters and headers only by alias; the secret never sits in the endpoint.
- Live-test an endpoint with a symbol, see the raw response next to the mapped example and map the response onto your own fields.
- Build data blocks: tick what you need in the response tree and give it a placeholder. In the prompt editor, hovering over the placeholder shows an example.
- As a manager, maintain sources, endpoints and data blocks. Prompts and API keys stay with admins.
The thinking behind it
The page was called “topic planning” and sat with the editorial tools. What it actually did was something else: configure data sources. Now it is named after what it does and sits under setup.
Figures in an article should come from your data, not from a model's memory. That is why a data block is not free text but a selected excerpt of a real response, fetched fresh for every name.
Keys do not belong in configurations that get copied and shared. Endpoints only name the alias, and each installation stores its own key.
Improvements
Saving leaves endpoints intact
The source form and the endpoint area only sent the fields they displayed when saving. Headers, mapping, parameter rules or notes could get lost that way, and for sources with many endpoints the first one was overwritten. Now saving only changes what you edited. Sources with several endpoints link to the endpoint area.
Only fields that do something
The “Queries” tab, tonality and system prompt on the endpoint, and the “Website” source type were read by no run and are gone. Existing website sources stay in the list.
Duplicate with endpoints
A copied API source takes its endpoints along. Duplicating is no longer offered for RSS feeds, because two sources with the same feed URL are not allowed.
New feeds join the scan straight away
Newly added RSS feeds are included from the next automatic scan; before, the scan worked through a fixed list and missed them. Dead feeds have been cleared out.
Manager role
Managers maintain sources, endpoints and data blocks without access to the prompts.
Named honestly and filed correctly
“Topic planning” became “Sources and endpoints”, and the surface now sits under setup. The intro says what the place does.
One endpoint editor fewer
An extra editor that could do less than the endpoint area in the settings has been removed. In its place is a link there.
Mutual links
The source page links to an endpoint's technical configuration, the endpoint area links back to source maintenance.
SEC filings endpoints ready
Endpoints for company filings with the US Securities and Exchange Commission are ready to test, with keys as an alias rather than a copied secret.
Data blocks
Named excerpts of an endpoint response that go into system prompts as placeholders.
Sources with API endpoints
Sources with their own endpoints and typed parameters.
Give feedback or ask a question
Missing something, bothered by something, unsure about something? Write it down here — it goes straight to the people working on this feature.
Or write to us right here: