Skip to navigation
MCP Server

What You Can Ask

Once the hosted MCP is connected, just ask in plain English. Your assistant picks the right KINETK tools for you.

Always mention “kinetk” in your prompt, for example “Use kinetk to…”. Without it, the assistant answers from its own general knowledge and skips KINETK.

“Use kinetk to show what’s trending in fitness this week.”

“What narratives is kinetk tracking about smartwatches?”

Free. Instant answers from the narratives KINETK already tracks.

Dig into a narrative

“Drill into the top narrative. Who’s driving it, and which platforms is it on?”

“Is that narrative growing or fading?”

Free. Works on any narrative the assistant found in a previous answer.

Get a full analysis

“Use kinetk to analyze the luxury watch conversation over the last 30 days, with written takeaways.”

Paid. Runs an analysis over up to 3,000 posts by default: about 45 credits, plus 1 credit for written takeaways. It can take a minute or two. Ask for a smaller analysis (minimum 750 posts) to spend less.

See what content looks like

“With kinetk, find what cozy home office content looks like.”

10 credits to match your question to the visual pattern map.

“What other content looks like that, even if it’s tagged differently?”

Free. Follow-up questions about patterns you’ve already found don’t cost anything.

Pull posts to cite

“Pull 1,000 recent posts about sustainable sneakers from kinetk so I can cite them.”

0.01 credits per post, so 1,000 posts cost 10 credits.

Privacy

Creator handles, display names and platform account ids are never returned. A creator only appears as a numeric id and a follower count.

See Pricing & Credits for how billing works. The MCP bills the same account at the same prices as the API.

Tool reference (for developers)

You don’t need these to use KINETK. They’re useful if you’re debugging an integration or checking which API call a tool makes.

Context jobs: live analysis (paid)

ToolDoesAPI call
kinetk_get_contextSubmits a job and waits up to about 24 seconds for the result. If it’s still running, returns { status: "pending", jobId }.POST + GET /intelligence/jobs
kinetk_create_context_jobSubmits a job and returns a jobId without waiting.POST /intelligence/jobs
kinetk_get_context_job_statusChecks a job’s status.GET /intelligence/jobs/{id}
kinetk_get_context_job_resultGets a finished job’s result: a short, token-efficient version by default, or the full payload with verbose: true.GET /intelligence/jobs/{id}

The job tools take kind: "records" or kind: "insights", with the same inputs as Intelligence Jobs. Through the MCP, limit must be 750–10,000, and window defaults to all.

Narratives (free)

ToolAPI call
kinetk_trending_narrativesGET /narratives/trending
kinetk_search_narrativesGET /narratives/search
kinetk_get_narrativeGET /narratives/{id}
kinetk_get_narrative_trajectoryGET /narratives/{id}/trajectory

Patterns

ToolAPI callCost
kinetk_search_patternspatterns job10 credits
kinetk_list_patternsGET /intelligence/patternsfree
kinetk_get_patternGET /intelligence/patterns/{patternKey}free
kinetk_get_pattern_contentGET /intelligence/patterns/{patternKey}/contentfree
kinetk_get_pattern_neighborsGET /intelligence/patterns/{patternKey}/lookalikesfree

ChatGPT deep research

ToolDoes
searchFinds matching content and returns { id, title, url }. Runs a 100-post records job, up to 1 credit per call.
fetchReturns one result’s full text and metadata by id.

ChatGPT’s deep research mode requires these exact names, so they’re the only tools without the kinetk_ prefix. Other assistants can ignore them.