MCP

Datasets

Upload a CSV through a constrained transfer ticket, or fall back to the Everyn Datasets page.

The public production MCP endpoint is live at https://mcp.everyn.ai/mcp. Configure a host with that Streamable HTTP URL and complete WorkOS OAuth.

A Dataset is an immutable, organization-scoped source snapshot. MCP never treats a local filesystem path as a hosted file, and it never fetches an arbitrary URL.

Use search_datasets and get_dataset after discovery. Search returns bounded receipts, not rows or full profiles. get_dataset returns metadata, schema/profile, up to five representative rows, and readiness. It never returns the complete CSV.

Upload when the client can move bytes

Call upload_csv with client_can_transfer: true, the filename, byte size, text/csv content type, sha256: digest, and an idempotency key.

The server issues a constrained transfer ticket and a signed transfer descriptor. The client PUTs the CSV to that descriptor.

LimitValue
Maximum size25 MiB (26214400 bytes)
Initial ticket lifetime15 minutes
Successful usesOne
RefreshRepeat the same unfinished upload_id to refresh without creating another Dataset

Observe intake with upload_csv and upload_id until a dataset_id is ready or the failure is terminal.

Fallback when the client cannot move bytes

If the host cannot transfer local bytes, call upload_csv with client_can_transfer: false. MCP returns a manual handoff to https://app.everyn.ai/datasets.

The user uploads in the Everyn app. The generic Datasets page does not complete the MCP ticket. After the upload is ready, rediscover it with search_datasets and continue from that dataset_id.

Job fit

Job fit is proven by sampling, not by a required-column compatibility check. Inspect representative rows, then run a sample before an unverified full Run.