← Back to Blog Guide

Log Time From Claude or ChatGPT in Five Minutes

Every time tracker now has an AI button. We went a different way. TimeQuorum doesn't add an assistant to the app; it lets the assistant you already use into your workspace, with your permissions and nothing more. This guide connects it in about five minutes.

Why we didn't bolt AI onto TimeQuorum

Most "AI time tracking" means one of two things: software that watches your screen and guesses what you were doing, or a chat box inside the app that only knows about that app. The first is surveillance with a friendlier name. The second is one more assistant to talk to.

What people actually want is to type "log 1.5 hours on the Acme redesign for yesterday" into the assistant that already has their calendar, their email, and their notes, and have it just happen. So TimeQuorum ships a hosted MCP server at your workspace address. Claude, ChatGPT, Cursor, Claude Code, and any other MCP client can connect to it directly. There is no plugin, nothing to run on your laptop, and no API token to paste anywhere.

Step 1: an admin turns it on

Agent access is off by default in every workspace. A workspace admin opens System Settings → Integrations and switches on the AI agents (MCP) card. The card shows the server URL, which is your workspace address followed by /mcp, and a table of every connected agent: who it belongs to, what it is allowed to do, when it last ran, and a Revoke button. MCP access is included with API access on Professional and Enterprise.

Step 2: connect from the assistant you use

Every client follows the same pattern. You give it the server URL, it registers itself with TimeQuorum, your browser opens a TimeQuorum consent page, you approve the permissions, and you're connected.

• Claude.ai: Settings → Connectors → Add custom connector, paste the URL.
• ChatGPT: Settings → Connectors (or Apps) → add a remote MCP server, paste the URL.
• Cursor: Settings → MCP → add a remote server, paste the URL.
• Claude Code: one command in your terminal:

claude mcp add --transport http timequorum https://acme.timequorum.com/mcp

Replace acme with your workspace name. Sign in if prompted, approve, done.

Step 3: approve only what you mean to

The consent page lists six permissions, and you can untick any of them. You can only grant what your own role in TimeQuorum already allows, so an agent connected by a Member can never see rates a Member can't see.

• Read time: your entries, time cards, and the running timer
• Write time: log and edit entries, start, pause, and stop the timer
• Read projects: clients, projects, classes, tasks, and services
• Write projects: create and edit clients, projects, classes, and tasks, and apply project templates
• Run reports: the reports you can already open
• Read billing: invoices and estimates

Agents never delete anything; they archive or retire instead. They can't send invoices, record payments, manage users, change settings, or approve time cards. Those stay in the app, on purpose.

Things to try

• "Log 1.5 hours on the Acme redesign for yesterday, design review"
• "Start a timer on the Northwind onboarding project"
• "What did my team log last week, by project?"
• "Which invoices for Acme are still unpaid?"
• "Create a project called Website redesign for Northwind, apply our Standard engagement template, and add a Kickoff task due Friday"

Behind each of those is one of the 40 tools the server exposes, grouped by the permission that unlocks them. The full list is in the MCP section of the docs.

What keeps this safe

An agent is a person's agent, not a system account. It signs in through OAuth 2.1 with PKCE, its access token expires after an hour and refreshes silently for 30 days, and after 30 days of silence it has to be approved again. Every call it makes is written to the audit log under the agent's name, next to the person it acts for. You can revoke your own agents from Profile → API Tokens; an admin can revoke anyone's from Integrations, or flip the workspace switch off and block every agent at once without deleting the connections.

The details are on our Trust & Security page, alongside everything else we say about how your data is handled.

Frequently asked questions

Does TimeQuorum have built-in AI?

No. There is no AI tier, and nothing in the app generates time entries by guessing. The MCP server lets the assistant you already use work with your workspace, with your permissions.

Can an AI agent see more than I can?

No. It acts as you, with your role, and can only be granted permissions you already have. If your admin hasn't turned agent access on, it can't connect at all.

Is there a per-call fee?

No. MCP access is included with API access on Professional and Enterprise. Your assistant's own subscription is between you and its vendor.

Does it work with assistants other than Claude and ChatGPT?

Yes. Any client that supports remote MCP servers over Streamable HTTP can connect: Cursor, Claude Code, and the growing list of desktop and IDE tools that speak the protocol.