FAQ
Common setup, troubleshooting, and source-selection questions for Cortex.
Use this page when you want the full set of frequently asked questions in one place, including source setup issues, connector tradeoffs, and common troubleshooting steps.
Choosing the right source type
Use these answers when you are deciding between RSS and more flexible but higher-maintenance source types.
When should I use Web instead?
Use Web when the site does not provide a usable RSS or Atom feed and you need Cortex to read content from a page directly.
When should I use RSS or API Endpoint instead?
Prefer RSS or API Endpoint whenever a stable feed or JSON API already exists.
Web is usually the right choice only when you need page scraping because the publisher does not expose a cleaner connector surface.
When should I use X List or X Search instead?
Use X List when the signal comes from a curated set of accounts, and X Search when the signal is defined by a query or topic rather than one timeline.
When should I use X Profile or X Search instead?
Use X Profile when one account matters most. Use X Search when the signal is defined by a topic query rather than a maintained set of accounts.
When should I use X List or X Profile instead?
Use X Profile when one account matters most. Use X List when the signal comes from a curated account set. Use X Search when the signal really is the query logic itself.
Working with RSS sources
These answers cover the most common setup and troubleshooting questions for RSS and Atom feeds in Cortex.
The source is active but nothing useful is appearing.
The feed may be valid, but too broad for what you want to track.
Try a more specific feed or lower the extraction cap so Cortex keeps a cleaner set of items.
The feed keeps surfacing old items.
Use Content Max Age when the feed republishes older entries or shows a long history that is not useful for your use case.
Working with Web sources
These answers cover the most common scraping and selector issues for Web sources in Cortex.
The source runs, but no items are appearing.
The listing selectors are probably not matching the current page structure.
Recheck the repeating container selector first, then confirm that the link selector actually points at a usable article URL inside each matched item.
The page clearly has content, but Cortex still misses it.
The page may render its cards after the initial load.
Add a small wait or a Wait For selector so Cortex does not extract before the listing page is ready.
When should I use RSS or API Endpoint instead?
Prefer RSS or API Endpoint whenever a stable feed or JSON API already exists.
Web is usually the right choice only when you need page scraping because the publisher does not expose a cleaner connector surface.
Working with X sources
These answers cover credentials, noise control, and connector choice across X Profile, X List, and X Search.
The source will not activate.
X sources require an active organization-scoped X API credential.
If the credential is missing or invalid, Cortex cannot poll the X API at all.
The timeline feels too noisy.
Start by excluding retweets, then decide whether replies are still useful for your workflow.
Only add engagement filters after that, because they can hide smaller but still important updates.
When should I use X List or X Search instead?
Use X List when the signal comes from a curated set of accounts, and X Search when the signal is defined by a query or topic rather than one timeline.
The source accepts the list URL, but the saved value looks different.
That is expected.
Cortex normalizes a full X list URL into the underlying numeric list identifier before storage.
The list is still too noisy.
First decide whether the list itself needs curation.
If the membership is already correct, then use reply, retweet, or light engagement filters to narrow the stream.
When should I use X Profile or X Search instead?
Use X Profile when one account matters most. Use X Search when the signal is defined by a topic query rather than a maintained set of accounts.
The query returns too much noise.
Tighten the query first.
Add operators, exclusions, or quoted phrases before leaning too hard on engagement filters, because a better query usually improves the source more than threshold tuning does.
What is the practical difference between relevancy and recency?
Relevancy follows X's ranking for what seems most relevant to the query.
Recency is better when you want a steadier capture of newer matching posts over repeated polls.
When should I use X List or X Profile instead?
Use X Profile when one account matters most. Use X List when the signal comes from a curated account set. Use X Search when the signal really is the query logic itself.
Working with API Endpoint sources
These answers cover JSON extraction, authentication posture, and common item-shape problems for API endpoint sources.
The source runs, but no items are extracted.
The most common issue is a wrong Items Path.
Verify that it points at the array inside the actual JSON response shape, not at the whole response object.
The API responds, but Cortex still cannot use the items properly.
Recheck URL Path and Text Path first.
Without a stable canonical URL or usable main text field, the connector can technically poll the endpoint but still fail to produce a good item stream.
How should authenticated endpoints be configured?
Store the credential separately, then reference it with Bearer Credential Ref.
Keep any stable non-secret request metadata in Request Headers JSON rather than embedding raw secrets in the source itself.
Working with MCP access
These answers cover the most common setup and troubleshooting questions for connecting MCP clients to a Cortex workspace.
I updated the MCP config, but I still do not see Cortex tools.
Restart the MCP client after changing its config so it reloads the Cortex connection.
If the tools still do not appear, confirm that API_SPEC_URL points to your workspace's /api/openapi.json endpoint and that the workspace is reachable from your machine.
Why am I getting authentication errors?
Confirm that AUTH_TYPE is set to bearer and that AUTH_TOKEN is a valid Cortex API key starting with ctx_.
The key must belong to the same organization as the workspace you are connecting to and must not have been revoked.
Why are some Cortex operations missing from the MCP client?
MCP exposes the full customer-facing Cortex API. If an operation is missing, confirm your API key has the required permissions and that the MCP bridge is running the latest version.