iGPT

iGPT provides email intelligence for agents and business workflows. The publisher's API page describes turning messages and attachments into cited answers and structured outputs, rather than simply returning a list of matching emails. Its pricing page separately describes a chat workspace and organizational deployments.

Reconstruct the conversation before acting

The API workflow connects email sources, indexes messages and attachments, then combines retrieval, context shaping and reasoning in one request. Listed capabilities include semantic and keyword retrieval with filters, document and spreadsheet processing, and reconstruction of threads across participants and time.
Structured output can make extracted action items easier to pass into a downstream workflow. The homepage's example schema contains a title, owner and due date. This is a documented example, not evidence that the request was executed or that every date and responsible person will be inferred correctly.
The source says answers link back to messages and attachments. Use those citations to check an extracted decision before sending a reminder, updating a CRM or treating an apparent approval as final. The publisher also lists SDK, MCP and LangChain access routes.

Understand the permission and storage boundary

iGPT describes permission-aware retrieval that follows the connected user's mailbox scope. Its FAQ says data sources connect through OAuth and that disconnecting a source removes it from the index. This is a service with ongoing indexing, not a promise that the underlying messages are never stored or processed.
Zero-retention language is scoped to inference inputs, prompts and outputs. Keep that separate from the email index and any organizational audit configuration. The publisher lists cloud, hybrid and on-premises options; its pricing FAQ says hybrid keeps data in the customer's VPC while inference runs in the cloud. It is not equivalent to a fully on-premises deployment.

Compare API usage with chat subscriptions

API pricing is usage-based, with separate input and output token rates and reasoning tiers. Chat Free, Pro and Team plans have different allowances and administrative features; a chat subscription is not documented as unlimited API usage.
The captured pages use different context-engine labels in the sample request and current pricing table. Consult the current endpoint documentation before choosing a tier. Published speed and cost comparisons are vendor benchmarks, not our own tests or guaranteed response times.