MCP Server · Claude Code · Cursor

Your workspace,
inside your AI client

Connect Claude Code, Cursor or any MCP client to 2pm.space with one API key. Ask it to build an agent, fix a workflow, fill a Drive table or put a chart on a board — it does the work through the same operations the app uses, and only the ones your key allows.

Connected in three steps

A key, a snippet, a request.

1

Create an API key

Under Settings › API Keys & MCP, create a key and choose what it may do. A key can never hold more than the person who made it.

2

Paste the snippet

Copy the server block for Claude Code or Cursor — the endpoint, and your key in an X-API-Key header — into the client's config file.

3

Ask for the work

Tell the client what you want. It lists the workspace's tools, reads the playbook for the job, and calls them step by step.

The whole workspace, as tools

Not a read-only window — the same operations the app performs.

Any MCP client

A remote server over Streamable HTTP, so it works with Claude Code, Cursor and any other client that speaks MCP. The in-app setup page gives you the snippet for each.

Over 300 tools

Agents and their canvases, tools, skills, knowledge, guardrails and memory, Drive documents, tables, mind maps and storyboards, data sources and boards, brands, the content calendar and channels.

Limited to the key

The server lists only the tools the key's permissions allow, and a key acts as the member who created it. Give a key read permissions only, and it cannot change a thing.

See roles and permissions

Playbooks for the long jobs

Nineteen built-in guides — building an agent, setting up RAG, curating a data source, planning a content calendar — that the client reads before it starts.

Your Drive, from the editor

Create and edit documents, fill Drive tables, add nodes to a mind map or shots to a storyboard — and the changes are there in the app for your team.

Data and boards

Query a connected database through its semantic model, build a board and add charts to it, or tune the model's measures and relationships.

About BI dashboards

What the server exposes

The connection, the key, and the parts of the workspace a client can work in.

Connection

  • A remote MCP server over Streamable HTTP
  • One endpoint, shown on the setup page
  • Snippets for Claude Code and Cursor
  • Discoverable at /.well-known/ai-catalog.json

Keys & permissions

  • A workspace API key in the X-API-Key header
  • A key acts as the member who created it
  • Permissions no wider than its creator's
  • An optional expiry date
  • Only permitted tools are listed

Agents

  • Create, configure and delete agents
  • Edit the workflow canvas node by node
  • Assign tools, skills and knowledge
  • Save and restore canvas versions
  • Run a test and read the trace

Drive & content

  • Documents, folders and Drive tables
  • Mind maps — nodes, shapes and tables
  • Storyboards — scenes, shots and frames
  • Brands, their products and images
  • Content calendar slots

Data

  • Data sources, tables and scopes
  • The semantic model: cubes, measures, joins
  • Queries in CubeQL or SQL
  • BI boards, tabs, charts and filters

Channels & inbox

  • Channels and their settings
  • Conversations and their messages
  • Care agents and triggers
  • Guardrails and memory

Copy-paste between tabs vs. MCP

The same change, made two ways.

Without 2pm.space
With 2pm.space

Describe the app to your AI, then copy its answer back into the app by hand.

The client makes the change itself, through the workspace's own operations.

A full-access token pasted into a chat.

A key limited to what its creator may do — and the tool list stops there.

The client guesses how a multi-step setup works.

It reads the playbook for the job first.

Rebuilding the same agent by hand in every workspace.

Ask once, and the client builds it node by node.

MCP questions

What people ask before they connect a client.

Work in your workspace without leaving your editor

Create a key, paste the snippet, ask.