LabscoMCP early access ↗
MCP SERVER · OFFICIAL PROJECT

Dreamlit Workflow Drafts

Builds automation workflows from an outcome described in plain language, and tells the agent to read the guidance first.

by Dreamlit AI ↗Workflow Automation & iPaaS✓ Official

LABSCO SUMMARY

Tool descriptions that tell the agent which call comes first are underrated documentation.

A model handed eleven tools will pick one that looks relevant and skip the one that would have told it how to phrase the request. Dreamlit writes the ordering into the descriptions — read the guidance, find the project, then create the draft — which is the difference between a workflow that describes an outcome and one that describes a guess. Note that creation produces a draft rather than a live automation, which is the right default for something an agent wrote.

What it is

Dreamlit's endpoint, which lists 11 tools and identifies itself as Dreamlit MCP 0.3.0. The tool descriptions are written as instructions to the calling agent — get_status is marked as the first call for product and prompting guidance.

What you get

  • get_status — read-only, and documented as the first call: product guidance, prompting guidance and approval state
  • list_projects and list_workflows — find the project or workflow by name, description or trigger type when only a name is known
  • create_or_update_workflow — create a draft, or update an existing one, from an outcome-oriented description
  • And the rest of the eleven, covering inspection and the surrounding workflow state

Requirements

A Dreamlit account. Add the URL, then sign in and approve access to one workspace; clients that cannot complete OAuth can send a personal access token, created in the project's settings, as a Bearer header instead. Using the MCP server adds nothing to your Dreamlit bill.

Setup effort

Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen