Team & platformGuide

MCP server

Waiso is an MCP server. An AI client that speaks the Model Context Protocol can list your projects, write and run scripts, build flows, read runs, set up schedules and more — as you, and never further than the token you gave it allows. You can watch what it does live in the app while it works.

8 min readBefore this: Create a projectOpen in Waiso
What you will learn
  • Create an MCP token and choose its scope
  • Connect each supported client with the command the dialog gives you
  • Know what an agent can and cannot do through the tools
  • Follow an agent’s actions live and revoke its access

An agent with a key of its own

The server lives at <your Waiso address>/mcp and speaks MCP over streamable HTTP. Every request carries an MCP Access Token in the Authorization: Bearer … header. A token belongs to one person and acts as that person, so the agent can never do more than its owner may do in a project — and the token’s scope narrows that further.

ProfileGlobal EmailADMINMCP TokensGoogle appADMIN
Manage scoped credentials for your connected tools.New Token
Claude Code — laptopdocs-demo · Read + WriteLast used4 min agoStatusActive
Cursor2 projects · Read-onlyLast usedYesterdayStatusActive
VS Code — CIdocs-demo · Scripts, Flows onlyLast usedNeverStatusActive
Old desktopAll projects · Read-onlyLast usedSep 2StatusRevoked
Overview
3Active tokens4 minLast used
Connect a client
Create a token and the dialog hands you the exact command for Claude, Cursor or VS Code.MCP docs
Settings → MCP Tokens: each token with its scope summary, last use and status.
Scope settingWhat it decides
ProjectsWhich projects the token reaches; every other project does not exist for it.
Read-only / Read + WritePer project: whether the agent may only look, or also create, change and run.
TopicsOptionally only some areas inside a project: Scripts, Flows, Schedules, Variables, Integrations, Robots, API, Email.
See and create projectsLets the token list all your projects and create new ones. Without it, it only sees the projects turned on.

Create a token and connect a client

  1. Open MCP Tokens

    In Settings, open MCP Tokens and press Create first token (or the new-token button).

  2. Name it and choose its scope

    Give a name you will recognise later, such as Claude Code — laptop, an expiry (or Never expires), and turn on the projects it may reach with their permission.

  3. Copy the token once

    Press Generate Token. The token is shown once — Copy this token now — it will never be shown again.

  4. Paste the client’s command

    The same dialog shows the exact command or config for each client with the token already filled in. Pick your client’s tab and copy it.

These are the four snippets, with your address and token in place of the placeholders.

Claude Code
claude mcp add --transport http waiso https://waiso.example.com/mcp --header "Authorization: Bearer <token>"

What an agent can do

The tools cover the product: create and run scripts and flows, read runs and their logs, manage schedules, variables, API endpoints, integrations and e-mail templates, and use the newer features too — state_get / state_set for script state, alerts_list / alert_rule_create for alerts, and git_status / git_commit_push / git_pull for Git.

Every call is checked

Each tool call passes the same check before it runs: the project must be in the token’s scope, the permission must cover a write, the area must be one of the token’s topics, and the person who owns the token must have the access in that project. A refusal names what is missing, for example this token's scope for project … is read-only.

Follow along and revoke

  • While a client acts on your account, the top bar shows MCP active. Click it to follow along: the app opens what the agent is working on — Working on script, Working on flow, Running — and Leave stops following.
  • Each token shows when it was last used. A token never used reads Never used.
  • Deleting a token stops it at once: any client using it loses access on its next call.

Related topics