An S3 bucket is a great place to keep a contract and a terrible place to find one. The console lists keys. It does not read the files, and it will not tell you which PDF mentions the renewal you are chasing.
Connect AWS S3 to ChatGPT through CorpusIQ and the bucket becomes something you can question. ChatGPT reads the listing and the file content at the moment you ask, then answers with the bucket and key it used.
The problem with the current workflow
Finding one file means knowing where it lives. Legal has a bucket, finance has another, and reports land under prefixes that made sense to whoever created them. You open the console, click into a prefix, sort by last modified and hope the name says what is inside.
Searching by content is worse. Download a dozen PDFs, open each, search each. So questions like "what was added this month" or "which files mention this customer" either wait for an engineer or never get asked.
How CorpusIQ solves it
CorpusIQ exposes S3 to ChatGPT through the Model Context Protocol, an open standard that lets an AI assistant call external tools. ChatGPT can list objects, read metadata and extract text from supported file types, all inside the buckets and prefixes your IAM role grants.
Read-only. ChatGPT cannot put, delete or modify anything, and cannot touch bucket policies or IAM roles.
Scoped. It sees only the buckets you grant. Nothing outside that trust scope is readable.
Live. Each question reads S3 when you ask.
What to ask
Paste these into ChatGPT once S3 is connected.
- "Which files were added to the reports bucket this month?"
- "Find the most recent contract uploaded to the legal bucket."
- "Show me invoices in the finance bucket from Q3."
- "Search every PDF across my buckets for the term ACME renewal."
- "List files larger than 100MB in the media bucket."
- "Pull the latest meeting notes from the operations bucket."
- "Which objects in the finance bucket are missing server-side encryption?"
- "Which tags are set on the files in the legal bucket?"
Follow-ups work. Ask for the month's additions, pick one prefix, then ask for the newest file in it and a summary of what it says.
Example report: new files by prefix
This is the shape of answer you get for the first question. The numbers are illustrative.
Which files were added to the reports bucket this month?
- 38new filesadded this month
- 1.4 GBcombined sizeacross all prefixes
- 5prefixes with additionsgrouped by key prefix
What to act on: The monthly-close prefix took a third of the additions. Check that the final versions are the ones you expect before the review.
Source: AWS S3, read-only. Numbers are illustrative.
Pair it with a CorpusIQ skill
The connector answers questions. Skills turn the answers into a routine. The Contract Intelligence Suite skill finds and compares contracts across connected file sources, so the latest version in a bucket can be set against the one you signed.
Setup in 3 minutes
- Sign up at corpusiq.io. Solo $29.95/month, 30-day free trial.
- In the CorpusIQ dashboard, click Connect next to AWS S3 and grant the read-only IAM role access to the buckets and prefixes you choose.
- Add CorpusIQ to ChatGPT from the dashboard setup page.
Where this earns its keep
Buckets that grew by habit get the most from it. Contracts in one place, invoices in another, exports nobody catalogued. A weekly "what was added" question replaces the walk through the console, and a content search replaces the download pile.
It also helps the people who never had console access. They ask in plain English and get the file, the bucket and the key.
What to watch out for
File content is read only for the supported types: PDF, DOCX, TXT, MD, CSV, XLSX, JSON and HTML. Other types show up as metadata. The answer is only as broad as your IAM grant, so a missing file may simply sit outside the prefixes you granted. Check the cited key before you act on a match.
See also
FAQ
Does CorpusIQ store my S3 files?
Direct MCP does not retain raw customer files or full connector response payloads. Operational logs may persist for up to 30 days. CorpusIQ does not train models on customer data; the selected AI client's plan and settings govern conversation handling.
Can ChatGPT upload, change or delete objects in my buckets?
No. The IAM role is read-only by design. ChatGPT cannot put or delete objects, change bucket policies or ACLs, edit IAM roles or read buckets outside the access you grant.
Which S3 data can ChatGPT see?
Object metadata such as key, size, last modified and content type, bucket listings for the prefixes you grant, encryption status and tags, and file content at query time for PDF, DOCX, TXT, MD, CSV, XLSX, JSON and HTML.
What plan do I need?
Solo at $29.95/mo includes AWS S3 and all 40+ connectors, with a 30-day free trial.