After an AI Agent Nearly Leaked Another Tenant's Revenue, Dev Open-Sources Canonic MCP Server
canonic-app · reddit · 2026-09-06
A developer found their multi-tenant analytics LLM agent generated technically valid SQL that joined across tenant boundaries — no prompt injection needed. Semantic layers like dbt describe metrics but don't enforce mandatory filters or tenant scoping.
They built canonic, an open-source context layer between agent and warehouse that compiles business rules (mandatory filters, required dimensions, tenant/role scoping) directly into every query, and returns answers with a trust score explaining why they should be believed. Ships as an MCP server for Claude Code, Cursor and others; ingests dbt/LookML or raw schemas.
More from coding & agent
- SurfSense: an open-source local NotebookLM with MCP agents and a prompt-caching lesson — FurtiveMirth · 2026-10-23
- procedural-memory: let agents replay saved tool-call sequences instead of re-solving tasks — Sad-Response1560 · 2026-10-23
- Med student seeking whole-library book search: Cognee's graph memory beats plain RAG so far — shrike293103 · 2026-10-23
- Dev demos AI models handling 3D objects and animations surprisingly well — chongdashu · 2026-10-23
- Jason Kneen burns 104.56B tokens and $24.7K in 304 days of AI coding on Codex and Claude Code — jasonkneen · 2026-10-23
- Typosquatted Blender MCP Repo Hit 958 Stars in Days and Silently Ran Malware — Promptmethus · 2026-10-23