MCP tools reference

The Neuralift MCP server exposes nine tools. All of them are read-only and scoped to the organization of the API key making the request. Seven return results from published use cases only; the two data-dictionary tools (list_prepared_datasets and list_prepared_dataset_columns) return organization-scoped metadata regardless of publication state. You rarely call these directly (your assistant picks the right tool from your question), but knowing what each returns helps you ask better questions.

Tools

Tool	What it returns
list_data_runs	Your organization’s published use cases: name, the goal and business context from the use case definition (products, channels, temporal context, segment size), start and completion times, and result-delivery statuses.
list_segments	The segments in your published runs: each segment’s title and summary, and, loaded automatically with each segment, its persona, insights, segment tactics, defining features, and KPI counts. Filter by use case (data_run_id) to focus on one run.
list_actions	The run-level recommended actions: rank, title, the impact-led strategy, the why-it-works details, rationale, mechanism, reach with its per-segment breakdown, and scores for strategic fit, expected impact, and executability. See Actions & action plans.
list_action_plans	The action plans under each action: the plan’s title, its angle, its targeted segments and contributing segment tactics, and the activation details (audience filter, SQL, and audience row count) for building the matching audience in your own systems.
list_segment_tactics	Segment tactics: the tactic, supporting details, the primary and secondary KPI it targets, the evidence behind it, and scores for evidence strength, expected impact, and executability.
list_insights	Segment insights: the insight, supporting details, a confidence score, the reasoning trail behind it, and its validation flag. See segment detail.
list_personas	Each segment’s persona: a first-person greeting, the persona narrative, and the defining features it’s grounded in.
list_prepared_datasets	The prepared datasets behind your analyses, the data dictionary at table level: table name, description, row count, status, and source type.
list_prepared_dataset_columns	The data dictionary at column level: column name and type, description, narrative, unique-value count, annotations, sample values, and summary statistics. Filter by prepared_dataset_id to focus on one table.

Shared query parameters

Every tool accepts the same optional parameters. Your assistant fills these in automatically based on how you phrase your question.

Parameter	What it does
filter	Narrows results to records matching field conditions: for example, only segments belonging to one use case, or only insights above a confidence threshold.
sort	Orders results by one or more fields, ascending or descending; for example, segment tactics by expected impact.
limit	Caps how many records come back, useful for “top 5” questions.
offset	Skips past earlier records, so the assistant can page through long lists.
result_type	Controls the shape of the response the assistant receives back.

Worked example

Ask your assistant:

What are the five strongest insights across my latest use case?

A typical resolution: the assistant calls list_data_runs sorted by completion date with limit 1 to find the latest published run, then calls list_insights with a filter on that run’s segments, a sort on confidence descending, and limit 5. It then summarizes the five insights it gets back, each with its confidence score and supporting details.

Because filtering, sorting, and paging happen on the server, the assistant retrieves only what your question needs rather than pulling everything and sifting locally.

Next steps

• Prompts and workflows: example questions for each tool
• Segments overview: what segments, insights, and segment tactics mean
• Data dictionary: the datasets behind the answers