AI agents that work as a team

Describe the team you want. RacLink sets up a department with a manager and AI agents, each with its own job and tools. The manager plans the work, the team does it, and every action gets a receipt.

AI Workspace — Your workforce
Employees
8
Budget used
41%
within limits
Active now
3
Revenue · Ada (Manager)
Ir
Iris
Revenue Researcher
62% of budget
Mi
Milo
Outreach Specialist
28% of budget

AI Workspace — inside RacLink

A team of AI agents with a manager

What is an AI agent team in RacLink?

AI Employees is RacLink’s feature for running AI agents together in departments. Each department has one manager and a roster of agents. Every agent has a name, a job and the set of tools it is allowed to use, and those tools run on the AI model chosen for each one, so each part of the job gets a model suited to it.

You write a request into the department’s chat and the manager decides who answers. Work streams live, agents can open threads to sort out a sub-task, and under each message the platform writes a receipt of what actually ran. If a reply claims an action that did not happen, it is corrected before you see it. With FinCodes, agents can also write code on your own computer, in the folders you connect, with every change approved by you.

Best for: projects with several steps, like finding leads, writing the outreach and booking the calls.

RacLink puts every top AI model in one workspace, including Claude, GPT and Gemini. See which AI model is best for which task.

In the app: AI Workspace
Capabilities

What AI agent teams does

A department turns one request into finished work: the manager plans it and each agent does the part its tools allow.

A manager that plans and checks

Each department has a manager who plans the work, picks who does each step, checks the result and closes it.

Agents with real tools

An agent can only take the actions its tools allow: send email, hunt leads, write documents, update CRM records.

Receipts, not claims

Under every message, “What was actually done” is written from the run record: a green tick, a red cross, or an amber clock for work waiting on approval.

Set up by conversation

Describe the team and the designer proposes the department, manager, agents and tools. Nothing is created until you accept.

How it works

How AI agent teams works

  1. Step 1

    Describe the team

    “A sales team that finds leads, writes the outreach and books the calls.” The designer proposes who does what.

  2. Step 2

    Choose how much it does alone

    Pick the department’s approval mode, then accept. New agents introduce themselves in the room.

  3. Step 3

    Write your request

    The manager assigns the work and you watch each step, or hand a single task over and follow it on the Tasks board.

Safeguards

You stay in control

You decide how much AI agent teams does without you. These are the controls that apply to it today.

Approval mode at every level

Set it per agent, per department or for the whole workspace: no approval, ask me first, or the manager approves. Nothing set anywhere means approval is required.

A budget for each agent

Give each agent a monthly limit. Hit it, and the work stops.

Only the tools you give

Each agent can use only the agents and tools assigned to it. Managers hold no tools themselves.

Stop at any time

Send becomes Stop while a run is going. Stopping parks the job where it stands and never rolls back work already done.

Compare allowances and credits on the RacLink pricing page, or connect Claude to your RacLink workspace with MCP.

FAQ

Questions about AI agent teams

What is a department of AI agents?

A department is a chat room with one manager and a roster of AI agents. You write a request, the manager plans it and assigns each step to the agent whose tools fit, and the team works in the background.

How is a team of agents different from one AI agent?

One agent handles one job. A department lets several agents split a project: the manager plans it, hands each step to the right agent, checks the result and reports back to you.

Can the agents send email without my approval?

Only if you allow it. Approval mode can be set per agent, per department or for the whole workspace. With “Ask me first”, anything outward-facing waits for you. If nothing is set, approval is required.

Which AI models do the agents use?

Each agent runs on the model set on its card, and the built-in agents let you change theirs. Every top model is built in, so you can pick any of them without an API key, and swap the model without the work changing hands.

How do I know the work was actually done?

Every message carries a receipt written by the platform from what actually ran, not from what the agent said. A claimed action that did not run is corrected before you see it.

Put AI agent teams to work

Create a free RacLink account to look around, then pick the plan that fits. Or book a live demo and see it run on your own work.