iGPT: your agent’s email context in one request

Email holds the context most agents are missing: decisions, commitments, attachments, the history of a deal. It is also one of the hardest sources to work with live. iGPT is a premium connector on Agent200 built for exactly that.

Why raw email is hard for agents

Working a mail API live means many calls: list the messages, fetch each one, follow the thread, open the attachments. Every payload lands in the model’s context window, raw, whether it matters or not. The agent spends its tokens reading, and you pay for every call along the way.

How iGPT works

1. Your end user connects their email to iGPT through OAuth.
2. iGPT indexes it automatically: messages, threads and attachments such as PDFs, documents and spreadsheets.
3. Your agent asks a question.
4. iGPT answers from its index in one request, returning the relevant context, structured, with citations to the source messages and files.

Because the source is already indexed, the agent makes fewer calls and spends fewer tokens than it would working the raw API.

What your agent gets

• Cited answers. Every piece of context points back to the message or file it came from.
• Thread reconstruction. Conversations arrive as conversations, not as loose messages.
• Attachment processing. The contents of PDFs, documents and spreadsheets are part of the answer.
• Permission-aware retrieval. Each user’s answers come only from what that user connected.

Same key, same bill

iGPT is a paid service billed through Agent200, like an AI model or web search. It uses your Agent200 API key and appears on the same statement. There is no separate iGPT account and no separate key.

Built on the same controls

Access is OAuth only, every user’s data is isolated from every other user’s, and the data is never used to train models.

See how an answer comes back, with its sources, on the iGPT page. The benchmarks against a live mail agent are there too.