A built-in AI agent is one that every project room has from the moment it is created. Nobody has to configure it. It has a handle, a field it knows well and instructions written for that field, and it says nothing until somebody writes its handle — @seo, @google-ads, @legal.
Who they are
- @google-ads, @ad-grants, @google-ads-dev and @shopping-feed: Google Ads accounts, the Ad Grants rules, the Google Ads API and Merchant Center feeds.
- @tracking, @chatgpt-ads, @meta-ads and @linkedin-ads: conversion tracking, and the advertising platforms of OpenAI, Meta and LinkedIn.
- @seo, @content, @ai-dev, @ai-builds and @legal: technical SEO, copy, building with AI tools, retrieval and function calling, and what each platform's own rules permit.
- @topvoice and @warmlike: two ready-made LinkedIn services, both working through your organization's top-voice.ai connection.
How they behave
They answer only when @mentioned, never on their own. They run on your organization's AI key, so a question to @seo costs exactly what a question to an agent you built would cost. A member can switch any of them off in one room by clicking its chip, and if you add your own agent with the same handle, yours answers instead.
Most of them advise rather than act. They do not hold your LinkedIn key and cannot send a message or publish a post. The exceptions are explicit. The Top Voice AI agent works through a connected top-voice.ai account. The Google Ads family reads the Google Ads accounts a partner has given your organization, read-only, except that the Ad Grants agent may propose changes which a member then applies.
Questions people ask
Can I change what a built-in AI agent says?
Not its instructions. Add your own AI agent with the same handle and your own instructions; in that room it takes the handle over, and the built-in one steps aside.
Which model does a built-in AI agent use?
With an Anthropic key, Anthropic's current default model. With an OpenAI or xAI key, the model you last named for one of your own AI agents, because a guessed model name there would fail on the first message.
Nothing happened when a built-in AI agent mentioned another one. Why?
On the default settings, a mention made by a built-in AI agent waits for a member to press Let them answer. The rule is under Settings, above the room list, set separately for each kind of agent.
Do built-in AI agents exist outside a room?
No. They live in project rooms. Outside a room, the closest thing is your own assistant connected to MakeHybrid over MCP, which has the LinkedIn tools but none of these specialists' instructions.