The problem MCP solves
A language model can write and reason, but it doesn’t know anything about your records. It can’t look up a client, schedule a meeting, or track hours. For that, it needs to connect to other software.
Before MCP, every AI app had to build its own integration for each program. MCP changes that: a program builds an MCP server once, and any assistant that supports MCP can use it.
Think of it like a power outlet: the appliance and the grid don’t need to know each other, as long as they use the same plug.
How MCP works
There are two sides. The server is part of the program that holds the data—like a time tracker. The client is in the AI assistant: Claude Desktop, Claude Code, ChatGPT, Cursor, Codex, or VS Code.
The server lists the tools it offers, each with a name, a plain-English description, and the fields it needs. The assistant reads those descriptions and picks the tool that matches what you ask.
- You add an MCP server in your assistant using an address, like
https://app.half5.com/mcp. - The assistant opens that program’s sign-in screen. You sign in and give permission (OAuth). Your password never goes to the assistant.
- You ask something in plain English, like ‘log two hours for the login fix’.
- The assistant uses the right tool and shows you the result.
Tools, not full access
An MCP server doesn’t give the assistant open access to your database. It only offers the tools the developer built, with the same rights as the account you signed in with. If your account can’t see coworkers’ hours, neither can the assistant.
You can revoke access just like with any connected app: in the program itself, or by removing the server from your assistant.
- Local or remote. An MCP server can run on your own computer or in the cloud. A cloud server doesn’t need to be installed: just paste in the address and sign in.
- Open standard. Anthropic published MCP as an open specification in November 2024; OpenAI, Cursor, and Microsoft support it too.
- Consent per connection. With a remote server, you sign in using OAuth, the same method as ‘Sign in with Google’.
Where it helps
MCP is most useful for work you already do in an assistant. A developer fixing a bug with Claude Code can ask the same assistant to log the time afterward. A consultant writing a report in ChatGPT can ask what’s already been logged this week.
It’s less suited for tasks where you want to check every field before saving. So choose software that keeps everything an agent does as pending or requires approval. How that works for hours is explained in Logging hours with Claude, ChatGPT, or Cursor.
How Half5 does it
Half5 runs its own MCP server at https://app.half5.com/mcp. Add that address in your assistant, sign in with your Half5 account, and give permission. There’s nothing to install.
The server offers six tools: log_work, log_work_batch (multiple entries at once), search_customers_projects, list_entries, update_entry, and delete_entry (only for your own entries that aren’t approved yet). A connected agent only sees your own hours, even if your account is a manager or admin; invoices, coworkers, and settings stay out of reach. Entries from an agent come in as ‘Not approved yet’ and otherwise work like regular time entries: they show up in your overview and go through approval and invoicing like your other hours. How log_work picks the right client is explained in How log_work works.
Frequently asked questions
Is MCP the same as an API?
Almost. An MCP server usually uses a regular API behind the scenes, but it describes its tools so a language model can pick and use them on its own. You don’t have to write code to let an assistant use it.
Which AI assistants support MCP?
Among others: Claude (Claude.ai, Claude Desktop, and Claude Code), ChatGPT through connectors, Cursor, OpenAI Codex, and VS Code. Where you add a server depends on the app.
Can an assistant change things through MCP on its own?
Only through the tools the server offers, and only with your account’s rights. Good software keeps an agent’s changes pending for review, so you or your manager can still approve them.
Do I need to install anything?
Not for a remote MCP server: just add the address and sign in. Only local MCP servers, which run on your own computer, need to be installed.
