Skip to main content
The Weavable MCP server exposes three tools. All tools require an authenticated session — see Connect a client for setup. The typical workflow is:
  1. Call list-contexts to discover what’s available and to obtain contextIds and insightIds.
  2. Call context-summary with a contextId, a natural-language query, and a time range to generate a written summary.
  3. Call context-insight with a contextId and an insightId from step 1 to generate a markdown insight (chart/table/visualization rendered as markdown).

list-contexts

Lists all contexts accessible by the authenticated user. Each entry describes the context’s data sources and the insight IDs that are compatible with it.

Parameters

None.

Returns

A JSON array of context objects: Contexts with no data sources are omitted. Results are sorted by most recently created first.

Example


context-summary

Generates a written summary from a context using a natural-language query and a time range.

Parameters

Returns

Markdown text containing the generated summary. While the summary is being generated, the server emits MCP notifications/progress events (rendered by the client as a running word count) so the UI stays responsive during long generations. If the request fails, the response is returned with isError: true and the text contains a human-readable error message.

Example

Request:
Response (text content, truncated):

context-insight

Generates a markdown insight (chart, table, or other visualization rendered as markdown) for a context using a given insight ID. Only insightIds returned by list-contexts for that context are valid.

Parameters

Returns

Markdown text with the rendered insight, prefixed with the insight title as an ## H2 heading. If the context cannot be found or the user lacks access, the response text is "Context not found or access denied". If the insight isn’t compatible with the context, the response explains that and points back to list-contexts.

Example

Request:
Response (text content, truncated):