SCALE MCP documentation

Version of 4 November 2026.

SCALE by split/build, provided by ApplePie creative agency OG (Vienna, Austria; see the legal notice), exposes your workspace to AI assistants through the Model Context Protocol (MCP). This page is for people who want to connect Claude, ChatGPT, Grok or another MCP client.

1Server

2Connect an assistant

Claude. Open Settings → Connectors → Add custom connector, paste the server URL, then sign in to SCALE and approve the permissions.

ChatGPT. Enable Developer mode, add a connector with the server URL and OAuth authentication, then sign in to SCALE and approve the permissions.

Grok, Cursor and other clients. Add an MCP server with the URL above; any client that supports OAuth 2.1 with dynamic client registration will start the sign-in itself.

On the consent screen you see which assistant is asking, choose the permissions (scopes) and the workspace, and can Deny. Denying returns access_denied to the assistant and gives it nothing.

3What the assistant can do

After connecting, call the get_started tool first: it returns the usage guide. This server offers a curated set of about 50 tools (the list your client shows is the complete list):

  • Projects and tasks — list, read, create, update, assign, comment, move, archive; subtasks and statuses.
  • Time and workload — log time, read time entries, read team workload.
  • Meetings — list, schedule, update and cancel meetings.
  • Knowledge base and files — read and edit pages, list attached materials.
  • CRM — companies, contacts, pipelines, leads and deals; moving a deal between stages.
  • Departments and your brief — list departments, get your own summary.

Every tool is annotated as read-only, destructive or open-world, so your client can ask before it changes or deletes anything. Archiving is marked destructive: it hides the item from lists, so your client asks first. Tools never return more than your own account may see. There are no tools for e-mail, Telegram, e-signature or third-party integrations on this server.

4Permissions (scopes)

You approve scopes on the consent screen, for example projects:read, tasks:read, tasks:write, comments:write, time:write, crm:read. A tool that needs a scope you did not approve is refused. Write scopes are not pre-selected beyond the recommended minimum. Your own rights in each workspace (roles and per-project access) limit every request further.

5Example requests

  • "List my open tasks due this week."
  • "Create a task 'Prepare the offer' in project Website redesign and assign it to me."
  • "Summarize the comments on task 42."
  • "Log 2 hours on the Website redesign project for yesterday."
  • "Which deals in the Sales pipeline are in the Negotiation stage?"

6Limits

  • Rate limits. Reads are not counted against the write budget. Writes are limited per token (by default 60 calls and 1,000 affected records per minute); when the limit is reached the tool returns rate_limited with retry_after_seconds.
  • Response size. Lists are paginated; use the returned cursor to continue. Very large responses are refused with a clear message instead of being cut.
  • Errors. Tools return readable errors (for example forbidden, resource_not_found, invalid_arguments), never internal details.

7Revoke access

  • In SCALE open Integrations → Connected apps and disconnect the assistant. It stops working immediately.
  • A workspace administrator can turn the connector off for the whole workspace in the same place.
  • You can also remove the connector in the assistant itself.

What an assistant already read stays in your conversation with it; delete it there. See the Privacy Policy, section 7.

8Support and legal

Questions and problems: support@splitbuild.net (1–3 business days).