Knowi Webhook MCP Server

Knowi Webhook

Knowi scheduled reports push dashboard and dataset exports on their upstream cadence — hourly, daily, weekday, or monthly. Each delivery is one stored report; columns, filenames, and shape depend on the dashboard or dataset the user scheduled. Tools read what already landed — there is no pull API and no way to trigger a new send from here.

What upstream deliveries look like

Knowi webhook schedules typically POST:

• Multipart FormData — separate parts for dashboard PDF (application/pdf) and dataset CSV (text/csv); one upstream fire may produce two stored reports — list recent deliveries when the user asks about “the PDF” vs “the CSV”.
• JSON-by-dataset — when the upstream format is JSON instead of a file attachment, expect an object keyed by dataset name whose values are row arrays; field names mirror the scheduled dataset columns.
• Custom POST bodies — conditional schedules may wrap a short default message or partial data instead of full tabular rows; inspect export.json or export.text rather than assuming CSV headers.

Walk export.rows for column-level KPI questions on CSV parts. For JSON dataset objects, read keys from the latest export.json before quoting metric names.

First moves

1. Latest report — default for KPIs, metrics, or “what did the scheduled export say”.
2. Report freshness — check source.captured_at and source.filename before quoting figures; Knowi fires on the schedule configured upstream, not on demand.
3. Recent reports — compare runs, spot duplicate filenames, or find a delivery from a specific date.
4. One report by id — after listing, when the user names a particular run.

Response shape

• source — capture time, filename, format, connection purpose.
• export.rows — parsed CSV preview when tabular; may truncate with rows_truncated.
• export.text — full body when rows truncate or the file is not tabular.
• export.json — parsed JSON when the stored payload was JSON.

Knowi may send multipart form data — separate parts for dashboard PDFs (application/pdf) and dataset CSVs (text/csv). When the upstream format is JSON instead of a file attachment, expect an object keyed by dataset name whose values are row arrays (field names match the scheduled dataset). Custom POST bodies may wrap conditional data or a short default message rather than full tabular rows — inspect export.json or export.text rather than assuming CSV headers.

PDF attachments store with limited tabular preview — prefer export.text or note binary content when the user asked about a dashboard PDF.

Operational order

1. Start with the latest report — do not guess a report id.
2. If captured_at is older than expected, say so — Knowi sends on the upstream schedule.
3. List recent reports only for history, duplicate suspicion, or explicit compare-by-date asks.
4. Read ## This connection in the skill supplement for property-specific column or field notes.

Gotchas

• Read-only — cannot trigger a new Knowi send or change report settings from here.
• Variable columns — do not assume fixed KPI names; read headers from the latest rows or connection notes.
• Format split — the same schedule may land CSV rows in one delivery and JSON in another if the upstream format changed.
• Multipart vs JSON — one POST may include both a PDF and a CSV; storage may surface them as separate captures or combined — list recent deliveries when the user asks about “the PDF” vs “the CSV”.
• Zip-wrapped CSV — unpacked automatically when a CSV is inside the archive.
• Empty — no stored reports means nothing has arrived yet; say no report is available.