Get started - free

Changelog

What we've shipped in Gurusup Brain, newest first.

  1. Agents

    Connect your agent with one card per tool

    Gurusup Brain is meant to be used from the AI tool you already work in. Connecting it used to be one generic step with a server address, and each tool wants that address in a different place.

    Now Connect your agent has a card for each client, with the exact steps for that tool: Claude Desktop, Claude Code, Codex, ChatGPT, Cursor, VS Code and Other MCP client.

    • Cursor and VS Code install in one click.
    • Claude Code and Codex give you one command to paste in your terminal.
    • ChatGPT adds it as a custom connector in Settings, and Claude Desktop as an MCP server in the desktop app.
    • Other MCP client gives you the server URL to paste into any client that supports remote MCP.

    Whoever sets it up only needs the card for their tool, and once it is connected they can ask about your company's knowledge without leaving the conversation they are already having.

    MCP reference →
  2. Memory

    Your first memory starts from the most useful content

    The first build used to wait until everything you imported had been read. With a large Slack history or a big Notion workspace, that meant waiting a long time before you could ask anything.

    Now Gurusup Brain starts with a first batch and brings in the rest on later passes. It reads uploaded files first, then Notion, HubSpot, webhooks and Slack, in that order.

    You can start asking much sooner, and the memory keeps filling in behind you. With very large imports, the first answers may not cover everything yet, and Ingest shows what is still being learned. As later passes finish, the same question gets a more complete answer without you doing anything.

  3. Sources

    Choose which Notion pages and HubSpot deals go into Gurusup Brain

    Not everything in a Notion workspace or a CRM belongs in your company memory. HubSpot used to offer only a Backfill button that brought in everything it could reach, with no say in what came in.

    Now Choose content lets you pick. Search, filter deals by pipeline, stage and date, tick what you want and press Import selected. In Notion, Choose content replaces Backfill.

    You can start with the pipeline or the pages that matter, for example this quarter's open deals and the sales playbook, and add more later. A focused memory gives more precise answers, and you stay in control of what goes into it. If you change your mind, open Choose content again and import more.

  4. Sources

    Webhooks tell you when content has arrived

    A webhook lets any tool send content to Gurusup Brain, even one we do not connect to yet. But after creating one there was no way to know whether anything had arrived, so you could not tell a working setup from a broken one.

    Now each webhook shows Waiting for content until the first delivery and Content received after it.

    • Generate new URL replaces a URL you can no longer see.
    • Admins can paste up to 1 MB of JSON with Send JSON, to try the webhook or send content without setting anything up outside.

    You can wire up an internal tool or an automation, send one test event and see straight away that it reached your memory, before you rely on it. If the status stays on Waiting for content, the problem is on the sending side.

  5. App

    Your organization, named the same way everywhere

    Gurusup Brain called the same thing a workspace in some screens and an organization in others, which made it unclear whether sign-up, plans and settings were talking about the same account.

    Now it is organization everywhere: sign-up, plans, settings and access notices. Team and Settings live in the sidebar under Organization, and any admin can rename the organization from Settings. Contradictions is now called Conflicts, and old links still work.

    One name for one thing means fewer questions when you invite your team or change your plan, and the word in the app matches the word in the docs. If you were used to Contradictions, everything that was there is now under Conflicts.

  6. App

    The docs are one click away, and your agent can read them too

    The sidebar now has Docs, and the screens where people get stuck (webhooks, file uploads, upload errors, Connect your agent) link to the right guide. The prompt you copy with Copy prompt also offers a read-only docs server your agent can use without signing in.

    Open the docs ↗
  7. App

    Inviting and managing your team works as it should

    Four bugs on the Team screen are fixed, and the screen itself now opens much faster.

    Fixes

    • Inviting someone who is already a member no longer holds a seat.
    • Resend and Cancel work again on pending invitations.
    • Re-inviting a waiting email no longer says it was sent when it wasn't.
    • Removing someone from the organization removes their access, even if they had a session open.
  8. Sources

    Sources shows what is really connected

    A round of fixes so Sources reflects what you have connected and what each source contributes.

    Fixes

    • Each source's share of your memory is now right. Slack and Notion used to show 0%.
    • An app whose authorization you left halfway no longer shows as connected.
    • Opening a HubSpot source with a single connection no longer leaves a blank screen.
    • Local uploads no longer offer a Remove source button that always failed.
    • Get started shows loading rows, and a Try again button if your apps fail to load.
    • ZIP and TAR.GZ files made on a Mac no longer process or charge for the metadata files macOS adds.
  9. Memory

    See what Brain is doing while it learns

    Empty screens used to leave new admins guessing. Now Ingest, Conflicts, Skills, Synapses, Queries and Sources show a preview of what they will hold until the memory is built, and a notice across the app says when it hasn't been built yet.

    • Each ingest shows one clear state (Queued, Learning, Learned, Paused or Failed) and a three-step progress bar. A failed ingest is no longer marked as learned.
  10. Agents

    Two new tools let your agent tell us when Brain falls short

    The agent using Brain had no way to tell us what went wrong. Two new MCP tools fix that:

    • report_issue: a failure the agent can point to. A tool error, information the user says exists but Brain didn't find, or an answer that contradicts its own evidence.
    • send_feedback: an improvement when nothing failed, sent once at the end of the task.

    Reports are linked only to the caller's own activity, by id, so we can trace what happened without reading anything else in your organization.

    MCP reference →