1. Contact support
Email paul@pacificsoftwareventures.com for product, access, connector, privacy, or security support. Pacific Software Ventures monitors the mailbox. We answer access problems and suspected security issues before general product questions.
Include the app you were using, roughly when, and what you expected to happen. Remove confidential record text unless it is necessary to diagnose the issue.
Email support2. Connect and sign in
Accept your invitation, sign in to CRE Brain, and open Settings, then Connectors. Copy the connector’s address, add it as a custom connector in Claude or ChatGPT, and sign in when asked. The install steps say where each app keeps its connector settings. CRE Brain is invitation-only. Signing in is not enough on its own; your account has to hold a seat in your firm’s workspace.
One connection follows you across claude.ai, Claude Desktop, and Claude on mobile. ChatGPT Business and Enterprise workspaces add the same address as a connector. The connection and your sign-in are the whole setup.
3. Troubleshooting
- Sign-in loops, or you are told you are not authorized
- Check that you have accepted your invitation and that the assistant is connected to the same account you signed in with. After an account or seat change, disconnect and reconnect once.
- No CRE Brain tools
- Check that the connector is switched on for this workspace or conversation, then start a new conversation so the assistant picks the tools up.
- No results
- Ask the whole question in plain words, and name the property, person, lender, market, or decision you mean. Searching your chat history is not searching CRE Brain; an empty chat search does not mean the firm has no record.
- Wrong or stale result
- Tell the assistant what is wrong, and it records your verdict on that answer. If the fix is a rule about how your firm works, ask it to remember the rule. Anyone at the firm can read, change or remove those rules under Settings, then Firm memory.
- A cited source is missing or wrong
- Ask the assistant which document or finding the answer came from. If it cannot name one, or the source does not say what the answer claims, send support the question you asked and the time, and we will look into it.
- An AI agent’s draft is wrong or incomplete
- A person at your firm approves everything an AI agent prepares, so change the draft before you approve it. If the same mistake repeats, send support the name of the work and the time.
- Temporary failure
- Retry once. If it happens again, send the exact message you saw and the time. Do not paste internal credentials or private record content into the report.
4. Data controls
- Get a copy of the record: a workspace admin can email support to ask for an export.
- Correct an answer: tell the assistant what is wrong, and it records your verdict on that answer.
- Change or remove a rule: open Settings, then Firm memory. A rule you change or remove there changes how the assistant works for everyone at the firm.
- End or change access: ask your organization’s CRE Brain administrator, or contact support if administrator help is unavailable.
- Privacy request: email support with “CRE Brain privacy request” in the subject. We may verify your identity and organization before acting.
5. Security reports
If you believe an invite link, OAuth grant, account, or workspace has been exposed, email support promptly with “CRE Brain security” in the subject. Do not include the secret itself. State which credential type was affected, when you noticed, and whether you still have account access so we can shut off the exposed access first.
Good-faith vulnerability reports should include clear reproduction steps and the affected surface, avoid accessing another customer’s data, and stop testing once sensitive data or service disruption could occur. We will confirm the report with you and say when it is fixed.
6. What to include
A useful support request includes:
- which app you were working in: Claude, ChatGPT, or the CRE Brain dashboard;
- the approximate timestamp and time zone;
- the visible error or unexpected behavior, paraphrased if it contains firm content;
- the result you expected and whether the issue repeats in a new conversation; and
- a safe screenshot with names, record text, invite links, tokens, and account identifiers removed.
Never send: passwords, MFA codes, OAuth credentials, private invite links, payment-card numbers, government identifiers, or an entire confidential source document.