Integrations
Connect an AI assistant to Facets.
Facets runs an MCP server. Connect an AI assistant of your own and it can read your projects, and the idea boards you run that said so before they opened, so you can ask it about your work in your own words. If you allow it when you connect it, it can also add tasks to your projects and change them.
Where an owner or admin of your workspace allows it, and only if you tick it, it can also read and add to the records you keep about the people you manage. It never reads anything from a survey.
Which assistants
Claude and ChatGPT, through a custom connector, and programs that run on your own computer, such as Claude Code. Facets names Claude and ChatGPT itself on the page where you agree; a program on your computer is named as one, with the name it gave shown as its own claim.
What it reads
You choose one workspace, and either or both of these, each in the words the consent page uses:
- Read your projects in this workspace: their sections, tasks, dates, updates and schedule, and the names of the people on them. This permission only reads.
- Read the ideas everyone can see on your idea boards in this workspace that said, before they opened, that an assistant may read them. It is not told who wrote them; a name written inside an idea is read as written. This permission only reads.
A board is readable only if the person who runs it switched this on before it opened, and the board’s own page told everyone on it so before they wrote anything. A board that opened without it switched on can never be read. On a board where only its manager sees the ideas, or before a board reveals them, there is nothing to read.
An assistant can never read other people's projects, an idea board that did not say before it opened that an assistant may read it, who wrote an idea, comments on ideas, surveys or reviews or their results, or a person's record, such as goals, check-ins or assessments, with one exception. Only where a workspace's owner or admin allows it and you tick it, it can read the names of the people you manage and when their one-to-ones are and, under a tick of its own, what your open follow-ups for them and your points for next time say; never one-to-one notes, highlights, anything else in a person's record or anything they wrote.
In a workspace whose owner or admin has turned off “Model-written output” under Settings → Workspace, no assistant can be connected, and one already connected neither reads nor changes anything there until it is turned back on.
What it may change, only if you allow it
The page where you agree has a box of its own for this, under “What it may change”. It is not ticked for you; leave it unticked and the assistant will not add or change tasks. Ticked, it reads as:
- Add tasks to your projects in this workspace and change them: a task's title, description, status, priority, assignee, section, dates and estimate. It cannot delete or archive anything, add or remove people, or change a project's settings. No email or message is sent for a change it makes; the people on the project see it in Facets and in their regular digest. Every change is recorded and marked in the task's history.
It changes tasks in your name. The task's history says when an AI assistant last changed it, and each line a change of status, assignee, dates, title or section writes there names you and says it was made through an AI assistant, for the people on the project to see.
The people you manage, only if your workspace allows it
An owner or admin of the workspace decides first, under Settings → AI assistants, and it is off until they turn it on. Where it is on and you have the manager tools there, the page where you agree has a group of its own, under “The people you manage”. None of its boxes is ticked for you, and each reads as:
- Read the names of the people you manage in this workspace, when your one-to-ones with them are, and how many follow-ups of your own you have open for each, with the next due date. On its own this permission reads no words. This permission only reads.
- Read what is written in your open follow-ups for the people you manage, yours and the ones they own, and in the points for next time you typed or accepted from a wrap-up. These are private notes about named people. Some were proposed by an AI model from your notes or a transcript you pasted and accepted by you, and may repeat what a person said; some may have been written by a previous manager of that person. It cannot read one-to-one notes, goals, check-ins, assessments, highlights or anything a person wrote themselves. This permission only reads.
- Add a point for next time, or a follow-up of your own, for a person. Each is marked as added through an AI assistant. A point goes to no AI model unless you tick it when you prepare a one-to-one. A follow-up waits in Facets, on your own page for that person, until you keep or remove it, and until then nothing else reads it. It cannot add a follow-up for the person to do, change, complete or remove anything, or write anything the person sees. No email or message is sent.
Managers may connect an AI assistant to their People records. Off unless you turn it on. With it on, a manager may let an assistant of their own read the names of the people they manage and when their one-to-ones are; read what the manager's open follow-ups for those people say, the ones the person owns among them, and the manager's points for next time; and add a point or a follow-up of the manager's own. It can read nothing else in a person's record and nothing from a survey. What an assistant receives is handled by its maker under the manager's own agreement with that maker, not under Facets' terms. Turning this off stops every such assistant at once.
It never reads anything from a survey, a 360, a pulse, or an onboarding or exit survey: no answer, comment, result or count, and not who answered.
How consent works
When you add Facets to your assistant, it sends you to Facets to sign in. Facets then shows you a page that names the assistant, says where your answer will be sent, lists what it would read and, if it asks, what it could change, and asks which workspace. You can untick anything you do not want it to read, nothing it could change is ticked until you tick it, and nothing is connected until you press Allow.
Nothing about the people you manage is ticked for you either, and it is offered only in a workspace that allows it.
The connection ends if the assistant does not use it for 30 days, and after a year at most. You can disconnect it at any time under Settings → AI assistants, and it ends if you leave the workspace.
The owners and admins of the workspace can see that you have connected an assistant, and how many, and whether it may read or add to your People records. They cannot see which assistant, or what it read.
What it can never do, and signed in by Facets
An assistant can never delete or archive anything, add or remove people, change a project's settings, or change anything on an idea board. No email or message is sent for a change it makes; the people on the project see it in Facets and in their regular digest. Every change it makes to a task is recorded, counted on Settings → AI assistants, and marked in the task's history as made through an AI assistant. A point for next time or a follow-up it adds for a person you manage sends no email or message either: it is counted, marked as added through an AI assistant, and seen by that person's manager alone, and a follow-up waits until the manager keeps or removes it. It can never add a follow-up for the person to do, or change, complete or remove a follow-up or a point.
Facets runs the sign-in itself: the key it gives your assistant works only for this connection, only for what you allowed, and is refused anywhere else in Facets. Every read and every change is counted, and you can see when your assistant last read, and last changed something, and how many times, under Settings → AI assistants.
Facets sends the assistant’s maker nothing on its own: the assistant asks, and Facets answers. What the assistant receives is handled by its maker under your own agreement with them. The privacy notice and who can see what say the same.
How to connect and disconnect
In Claude or ChatGPT, add a custom connector and paste this address:
https://facets.team/api/mcp
Then sign in to Facets when your assistant asks, choose what it may read, and tick “What it may change” only if you want it to add and change tasks. To disconnect it, open Settings → AI assistants in Facets and press Disconnect; it stops reading and changing anything at once. To let an assistant that only reads change tasks, disconnect it and connect it again with the box ticked.