Back to the changelog

• Integration

  Document ingest from URL

  Automations point at a file URL (signed links included) and FlowSharp downloads it safely straight into storage: synchronously with an immediate response, or asynchronously with a queryable status and automatic retries. Passing an existing document id creates a new version, honoring the field's policy.

• Integration

  Discovery and download for automations

  With the same API key you can list a process's documents field by field and get a short-lived signed URL to download a file — no interface required.

• Feature

  Documents for AI agents

  Two new MCP tools: one to ingest a document from a URL into a flow, field or process (flow code supported), one to get the signed download link. Agents now handle documents end-to-end.

• Security

  Full defenses on URL downloads

  Https only, internal networks and service addresses blocked at every step (redirects included), field limits enforced before downloading, source URLs never logged. And a pending document can only be attached to a new process by whoever uploaded it.

• Design

  Document field refinements

  The expression picker suggests document properties only where templates actually resolve them; an empty multi-file field shows count 0 in task panels too; documents inside nested objects become clickable links in the data view.