Coming soon on 8frame. Agents that sit on the board as nodes and answer in the board chat like teammates are not released yet. This page explains the idea, how creative teams handle AI work today, what you can already do on the 8frame canvas, and what the agent features will do when they ship. We will not give a date until they are live.
Most creative teams already use AI every day. The trouble is where it happens: in one person's chat window, with one person's copy of the brief, producing results that someone then downloads and pastes into a deck. Microsoft's 2025 Work Trend Index found that 36% of leaders and 21% of employees expect to manage AI agents as part of their job within five years. If agents are going to be part of the team, the question is where they work and who can see what they do.
If you need the basics first, start with what an AI agent is. This page is about one specific shape: the agent as a teammate on a shared canvas.
The problem with private AI chats
A private chat is a fine place to think. It is a poor place to produce work a team has to agree on, for three reasons:
- Context is split. The brief is in one chat, the references in another, the client's notes in an email. Each AI conversation sees a different slice.
- Results are orphaned. A generated image leaves the chat as a file. Its prompt, its settings and the reasoning behind it stay behind.
- Nobody else can steer. A colleague who disagrees with a direction has to ask the person who owns the chat to rerun it.
We cover the context side in more depth in briefing AI with the same context as your team.
What works on 8frame today
The canvas is the part that exists now. On an 8frame board you place generation nodes (image, video and audio models), wire your own uploaded images into them as references, and keep every result on the board next to the prompt that made it. The board is real-time and multiplayer: collaborators see each other's cursors, and you can leave comments on the work.
A few facts that matter for teams, as of today:
- Inviting collaborators needs a Creator plan (collaboration up to 3 members) or an Agency plan (up to 30). Starter cannot add collaborators.
- Credits are not pooled. Each generation is charged to the person who runs it, on their own plan.
- There is no public share link for a board: access is the owner, invited collaborators (each with an 8frame account) and published templates.
- Failed generations are refunded automatically.
So a team can already work on one board, with every person's AI work visible to the others. What is missing is an AI that works on the board with you, rather than inside someone's chat.
What is coming: the Agent node
The agent features will put an AI teammate on the board in two places at once.
On the board, as a node. You will add an agent the same way you add anything else: from the «+» catalog in the bottom panel, from the right-click menu, or from the Library. The node's header will simply read «Agent» until you name it. Its name, avatar, role and standing instructions will live in the node's settings, so the same generic node can become a researcher, a producer or whatever role your team needs. The home page shows two examples, a Researcher and a Producer.
In the board chat, as a teammate. You will be able to type @ and the agent's name in the board chat, the way you would mention a colleague. The agent's replies and results will come back into the chat under its own name and avatar, with a way to jump to the result on the board. If you mention a role that is not on the board yet, the chat will offer to add it for you.
What an agent will read
An agent will read what is wired into it and nothing else: documents placed on the board, a brand, a selection of nodes you hand it, or another agent's results. The node will show what it reads, so everyone on the board can see the agent's context instead of guessing. Reading the web is planned as an option you allow per agent.
Where its work will go
When an agent produces something, it will place the result as nodes on the board, inside its own frame, marked as the agent's work. The same result will appear once in the chat. While it works, the agent is planned to show in the board's presence list and as a cursor, the way a human collaborator does.
What it will not do on its own
Anything that spends credits will wait behind an approval card: a short plan in plain words with Approve and Not now. Nothing will be generated before someone approves. We explain that flow in approving what an AI agent does before it spends credits. In the first version, an agent works when a person launches it or mentions it; there are no scheduled or unattended runs.
You will also be able to pause, stop, redo or remove an agent from its node or its thread.
A team workflow you can run today
You do not need agents to get most of the benefit of one shared board. A pattern that works now:
- Put the context on the board first. Upload the reference images and keep the written brief in a comment next to them, so every collaborator works from the same material.
- Give each direction its own area. One cluster of nodes per direction, each with its prompt visible. A colleague can add a node beside yours and try a variation without touching your prompt.
- Use cheap stills to decide. A Nano Banana 2 image is 11 credits at 1K (the node's default); Nano Banana Pro is 21 at 1K or 2K. Settle the direction with stills before anyone animates.
- Keep decisions in comments. Write the yes, and who gave it, on the board. There is no approval button or node lock today, so the comment is your record.
When the agent features ship, the same board becomes the agent's workspace: the brief you already placed is what it reads, and the directions it proposes land next to yours.
For a fuller picture of team setup, see the collaborative AI canvas for teams, and if you are comparing canvases, the best infinite canvas AI tools in 2026.
FAQ
Can I add an AI agent to an 8frame board today?
Not yet. Agent nodes and @-mentions in the board chat are coming soon. Today the board holds generation nodes, your uploads, comments and real-time collaboration.
Will an agent spend my credits without asking?
It is designed not to. Every step that spends credits will wait for someone to press Approve on an approval card.
What will an agent be able to read?
Only what is wired into it: documents on the board, a brand, a selection of nodes or another agent's results, and the web if you allow it. The node will show its inputs.
Will agents replace collaborators on the board?
No. They are planned as additional teammates on the same board. People still invite each other as they do today, and people approve what the agents propose.
Is there a release date?
No date is announced. This page will be updated when the features are live.
Sources
- Microsoft Work Trend Index 2025, "2025: The year the Frontier Firm is born" (36% of leaders and 21% of employees expect to manage agents within five years; 31,000 workers in 31 markets; published 2025-04-23): microsoft.com/en-us/worklab/work-trend-index/2025-the-year-the-frontier-firm-is-born, checked 2026-10-21
- 8frame prices, defaults and limits: the 8frame app, checked 2026-10-21
- Collaborator limits by plan: 8frame pricing, checked 2026-10-21
- Upcoming 8frame features: the 8frame product roadmap, 2026-10-21
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.