Answer the business question without writing a throwaway script.
Your database has the record. Your repo has the release and delivery data. Your CRM has the history, and the inbox has the promise someone made. CorpusIQ is the read-only layer that lets ChatGPT, Claude or Perplexity read all of it at once, so a product question is answered in a minute instead of a ticket. Answers carry the source they came from, so you can check the number yourself.
30-day free trial. No credit card required.
- Only reads your data. Never changes it.
- Checked by an outside auditor
- CorpusIQ does not train models on customer data
- Works with ChatGPT, Claude, Perplexity and Slack
Business questions arrive faster than anyone can answer them
- 01
A product question becomes a ticket and a two-day wait
"How many accounts on the new plan churned?" should take a minute. Instead it is a request in a queue, or a context switch you did not budget for.
- 02
Production data gets queried by hand, with no record of who asked why
A throwaway script against the primary database answers it once, then disappears. Nobody can see what was asked or what the number was based on.
- 03
The repo holds the schema. Nobody knows what the business calls it
You can read the tables. Knowing that this column is what finance calls "booked revenue" lives in someone's head or a thread from last year.
- 04
Debugging a customer issue spans four systems
The database has the record, the CRM has the history, the inbox has the promise that was made, and the checkout has what actually happened.
- 05
Onboarding is asking colleagues where things live
There is no single place that answers "where does this number come from" for the systems the business actually runs on.
What it can read
Access is scoped to the sources you authorize and answers come through read-only access. It does not commit, deploy or review code.
From question to cited answer
Eight examples, written the way an engineer would type them. The first five are engineering tasks. The last three cover website and search work, including answer engine optimization (AEO) and generative engine optimization (GEO). Both mean improving the content, schema and crawlability those engines read.
Blast radius before a deprecation
Situation. The team plans to retire the legacy export endpoint.
Ask. "Before we retire the legacy export endpoint, how many active accounts used it in the last 90 days, which plan are they on, and who owns each account?"
Reads, when the connector is connected: PostHogPostgreSQLHubSpot
What comes back. A count of active accounts, then a list grouped by plan with the account owner on each line. Every line cites the query or record it came from.
Then you. You decide the deprecation date and warn the right owners.
Incident triage after a release
Situation. Support volume spiked the day after a release.
Ask. "Support volume spiked on Tuesday. What shipped Monday, and did errors rise for any customer segment?"
Reads, when the connector is connected: GitHub (release and delivery data)PostHogSlack
What comes back. A timeline of the release, error counts by customer segment, and the support thread that first flagged it. Each item cites its source.
Then you. You open the code and fix it. CorpusIQ does not review the change.
What a column means in business terms
Situation. The team needs to know what a column means before anyone changes it.
Ask. "What does orders.status_v2 hold, which values are in use, and how does finance describe it?"
Reads, when the connector is connected: PostgreSQL (table description and distinct values)Google Drive or Google Sheets (the finance definition)
What comes back. The column description from the database, the values in use with counts, and the finance wording with the document it came from named.
Then you. You update the code or the docs with the agreed meaning.
One customer bug across systems
Situation. A customer reports being charged twice.
Ask. "A customer says they were charged twice. Show the checkout record, the Stripe charges and the email thread."
Reads, when the connector is connected: PostgreSQLStripeGmail
What comes back. Three matching records side by side with ids, so the engineer can see the order of events.
Then you. You reproduce and fix the bug.
Ground your own coding agent in production facts
Situation. A migration starts with a question about production data.
Ask. (Sent by your own MCP-capable coding agent through CorpusIQ.) "Before I write the backfill for status_v2, list the distinct values and row counts."
Reads, when the connector is connected: PostgreSQL
What comes back. The exact values and counts, cited to the query.
Then you. Your coding agent writes and tests the migration. CorpusIQ only supplied the facts.
Website and SEO fixes from search data
Situation. Pages are losing clicks and nobody knows which ones.
Ask. "Which pages lost clicks in the last 28 days compared with the previous 28, and what are the top queries for each?"
Reads, when the connector is connected: Search ConsoleGA4
What comes back. Pages ranked by click change, with the queries behind each one. Every line cites the report it came from.
Then you. Your coding agent rewrites the title and description for the worst pages and opens a draft pull request. After merge, ask again to confirm.
Answer engine content (AEO) from the questions people actually ask
Situation. People search in questions, and the site does not answer them directly.
Ask. "Which queries phrased as questions bring impressions but few clicks, and which pages target them?"
Reads, when the connector is connected: Search ConsoleAhrefs (keyword context)
What comes back. A list of question queries with the page currently ranking for each one, every line cited.
Then you. Your coding agent adds a concise answer block and matching FAQ schema to those pages in a draft pull request.
Crawl and index health (the base layer for AEO and GEO)
Situation. Some submitted URLs are missing from the index and nobody knows why.
Ask. "Which submitted URLs are not indexed, what reason does Google give for each, and how has Bing's crawl of the site changed over the last 28 days?"
Reads, when the connector is connected: Search Console (sitemaps and URL inspection)Bing Webmaster
What comes back. The unindexed URLs with the stated reason for each one, and the crawl trend. Each line cites the report it came from.
Then you. Your coding agent fixes the cause in the repo, for example a canonical, a noindex, a redirect or a sitemap entry, and opens a draft pull request. After merge, ask again to confirm the URLs moved.
Let your agent act on what CorpusIQ reads
CorpusIQ supplies the evidence from your live systems through read-only external-source retrieval. Your own coding agent makes the change in your repo, and a human merges it.
Read
Your agent asks CorpusIQ for evidence through a direct MCP connection.
Decide
The agent picks a change based on that evidence.
Act
The agent edits the repo with its own tools and opens a draft pull request.
Verify
After a human merges, the agent asks CorpusIQ again and compares.
What a cited answer shows you
SELECT plan, COUNT(DISTINCT account_id) AS accounts
FROM events
WHERE occurred_at >= CURRENT_DATE - INTERVAL '90 days'
GROUP BY plan
ORDER BY accounts DESC;Source: events table, via the PostgreSQL connector.
The query is read-only, and the answer names the table it read.
Connect your own agent
The same CorpusIQ account can also be reached over a direct MCP connection, from any MCP-capable client. Your agent works from the same cited evidence you see in chat, through read-only access. Setup steps and the login flow live on the direct connection page.
What changes in the week
- 01
Answer a product question without opening a ticket
Ask in plain language, get the number with the table and query it came from, and move on. No throwaway script, no queue, no context switch.
- 02
Describe a database column in business terms
The table description says what a column stores; a finance definition in Drive or Sheets says what the business calls it. CorpusIQ reads both and shows where each answer came from.
- 03
Debug a customer issue end to end
One question that spans the database record, the CRM history and the email thread, instead of three tabs and a guess about ordering.
- 04
Get onboarded from the systems themselves
New engineers ask the systems where things live, rather than interrupting the person who has been there longest.
Shopify
QuickBooks
Stripe
Meta Ads
Google Ads
Klaviyo
Google Analytics 4
HubSpot
Gmail
Slack
Google Drive
Outlook
Mailchimp
Etsy
TikTok
Instagram
Facebook
YouTube
Google Sheets
Google Docs
Google Calendar
Google Search Console
OneDrive
Dropbox
Calendly
PostHog
Semrush
Ahrefs
Axonaut
SHOPLINE
Airtable
PostgreSQL
SQL Server
AWS S3
MongoDB
Azure Cosmos DB
ShopifyRead-onlyQuickBooksRead-onlyMeta AdsRead-onlyGmailOffThis is how IT stays in control. Each tool is set to read-only or off. Read-only: AI can look at the numbers, never change them. Off: AI can't see that tool at all.
CorpusIQ can read your numbers. It can't change them.
What your IT person will ask, answered once.
- Only reads your dataCorpusIQ reads from your tools and never changes them. Nothing is written back.
- Checked by an outside auditorCASA Tier 2 certified by DEKRA. See the certificate.
- CorpusIQ does not train models on customer dataHow data is handled is written out on the security page.
- IT decides what AI can touchAdmins choose which tools AI can reach. Hosted on Microsoft Azure.
- The sample company connects nothingIt runs on sample data. Nothing is connected or sent.
Common questions
What engineering teams ask before they connect their production systems.
Does CorpusIQ write code or open pull requests?
No. Answers come through read-only access. It does not commit, deploy or review code: what it supplies is context read from the systems you authorize, returned with the source of each number. An agent that changes code is your own agent, acting on the evidence CorpusIQ reads.
Can it run SQL against production?
It reads through the connectors you authorize, including PostgreSQL, SQL Server and ODBC, and it is read-only by design. The alternative it replaces is a throwaway script on someone's laptop with no record of what was asked or where the number came from.
What does the GitHub connector read?
Release and delivery data: releases, pull requests, reviews, defects, churn, and GitHub Actions, all read-only. It provides release forecasts and code-quality trends. The GitHub connector cannot modify repository data, and it does not comment, commit, merge or review.
How is this different from just using ChatGPT?
ChatGPT alone cannot see your database, your CRM or your inbox. CorpusIQ is the read-only layer that lets it, with every answer citing the source it came from.
Do we have to give up control of the connection?
No. Access is scoped to the sources you authorize and can be revoked at any time.
Can an AI agent update my website using CorpusIQ?
Yes. Your agent reads the evidence through CorpusIQ, makes the change in your repo, and a human reviews the pull request. The write happens through your own tooling, and answers come through read-only access.
What does it cost?
Solo is $29.95/month and includes the 40+ connectors. There is a 30-day free trial and no credit card is required to start.
Read your systems in plain language
Start with one connection and one question. 30-day free trial, no credit card required.
Connector directory