Entagl MCP

MCP (Model Context Protocol) is a secure way to connect an app to an AI assistant. Connecting the Entagl MCP lets your assistant — Claude or OpenAI Codex — read and update your Entagl workspace on your behalf, and analyze or debug your AI's conversations when you ask: the same things the in-app Assistant does, from wherever you already work.

Ask questions like “what is my AI agent set up to do?” or “list my services and prices”, and Claude reads your current setup and answers in plain language — no change is made.

Tell Claude to add a FAQ, change a price, update your business hours, refine your agent's instructions, or create an offer, and it makes the change for you — reading the current details first so it fits what's already there.

Ask Claude to analyze your conversations — “what happened this week?”, “which customers dropped off after hearing the price?”, “why didn't it book that customer?”. Claude can find conversations by customer name, phone number or Instagram handle, read the transcripts, and inspect exactly what your AI did behind the scenes (which tools it used and why). It can then fix the problem with you — update a price, add a FAQ, adjust instructions — and test the fix, all in the same chat. Because conversations are untrusted customer text, Claude confirms any change with you before applying it after reading them.

There's no separate approval step — a change you ask Claude to make goes live in your workspace immediately. When you change how your AI agent behaves (its instructions), Entagl runs a quick safety check against your live setup first and only applies it if it passes. Claude tells you exactly what happened either way.

You sign in with your Entagl account, so the connection only ever touches your own workspace. Conversation content is shared with your AI assistant only when you explicitly ask Claude to read or analyze conversations, and every such read is recorded in an access log you can always account for. Connection secrets and credentials are never exposed.