Looped MCP
Work in Looped through an AI agent such as Claude, ChatGPT, Codex or Cursor, with your own permissions
Looped has an MCP server, so an AI agent can work in your Looped companies for you. You talk to Claude, ChatGPT, Codex or Cursor the way you'd talk to a bookkeeper, and the agent uses Looped's tools to do the work.
For example, you can ask it to:
- "Log three hours on the Acme website project for yesterday."
- "Invoice Acme for everything logged on their project this month, and send it."
- "Import this bank statement and propose accounts for every line."
- "Capture these four supplier invoices from my downloads folder."
- "Who owes us money, and how long has each invoice been outstanding?"
- "Draft October's pay run and tell me what changed since September."
- "Run month-end for Acme Trading."
The agent works as you
When you connect a client, you sign in to Looped and approve it. From then on the agent acts as you, with the role and permissions you have in each company. If your role can't approve statement lines in a company, the agent can't either, and every change it makes shows in the audit log under your name.
When you belong to more than one company, the agent asks which you mean before it acts. It never carries an action from one company over to another.
You decide what reaches the books
An agent prepares work and you decide how far it goes. Ask it to "just show me" and it only reads and drafts. Ask it to "do it" and it does everything that can be undone, then stops and asks before anything that can't be, such as emailing an invoice to a client. Some things, like closing a financial year or filing a return, stay with a person in the Looped web app whatever you ask. How far the agent goes has the full list.
Pages in this section
- Connecting an agent for setting up each client and the Looped skill.
- How far the agent goes for what it does on its own, what it asks first and what it never does.
- Playbooks for the step-by-step jobs the agent knows, from onboarding a company to month-end.
- Working well with an agent for memories, routines, files and what to do when something goes wrong.
Are you an AI? Visit llms.txt — these docs as plain markdown.