MCP support, Amazon Redshift, offline scheduled imports, Mermaid diagrams in AI chat, and a cleaner run details experience — all in one month.
May was a big month for Brooked. We shipped MCP (Model Context Protocol) support, added Amazon Redshift as a native connector, fixed a long-standing rough edge with scheduled imports and Google auth, and made the AI chat experience meaningfully better. Here's everything.
MCP integration — your data sources as AI tools
Brooked now ships a built-in MCP server. Connect any of your data sources — Snowflake, Redshift, Postgres, QuickBooks — and they become callable tools for any MCP-compatible AI client (Claude Desktop, Cursor, and others). No custom integration code needed.
The new MCP Connections tab in the add-on lets you generate and manage API keys, copy your server endpoint, and see which sources are exposed. Each source becomes a named tool the AI can call directly — the AI gets your live data, not a stale export.
- API key management — generate scoped keys, revoke them at any time, no shared credentials.
- Automatic tool naming — each data source gets a unique, collision-safe tool name. Multiple Google Sheets connections no longer crash the server on init.
- Streamable HTTP transport — the server speaks the latest MCP spec, with SSE fallback for older clients.
Amazon Redshift connector
Redshift is now a first-class connector. Query tables, views, and custom SQL, pull results into Google Sheets, and schedule refreshes — same as every other data source in Brooked. We also increased connection timeouts across all SQL connectors so large warehouse queries don't time out prematurely.
Offline access for scheduled imports
Scheduled imports run in the background — but Google's OAuth flow historically required you to be actively signed in when the job ran. We fixed that. Brooked now requests offline access during the Google auth flow, stores a refresh token, and uses it silently at run time. Your midnight sync no longer fails because your browser session expired.
As part of this, we improved the error message when a destination Google Sheet has been deleted. Instead of the raw API error ("Requested entity was not found"), you now see: The destination Google Sheet no longer exists. It may have been deleted from Google Drive. Please update the import to point to a valid spreadsheet.
Mermaid diagrams in AI chat
When the AI needs to explain a data relationship, a pipeline, or an entity structure, it can now render a Mermaid diagram inline in the chat response. Diagrams are generated server-side and rendered with unique IDs per stream so multiple charts in a conversation don't collide.
Run details slide-over
Import runs, export runs, and alert runs now have a slide-over detail panel. Click any run in the history list to see the full execution log, row counts, error details, and timing — without leaving the page. This replaces the previous modal that had limited space for error context.
What's next
We're working on smarter import auto-pause behaviour (the scheduler currently re-queues paused imports — that's being fixed), more connector types, and tighter AI chat integration with scheduled data. If you have a connector or workflow you'd like to see, let us know.
