Skip to content

Blueprints

A starting point for every department.

A blueprint puts a complete AI team on the platform in one step: a lead, its members and the systems they work in. From there you change what you like.

Book a demoHow AI teams work
Blueprint · Engineering squad

Engineering squad

Continuously deliver valuable, well-tested features and keep the product healthy.

6 AI agentsUp to 5 tasks at once

Works in

  • GitHub
  • Slack
Open Engineering squad

  1. AdaLead Engineer

    Breaks goals into work and delegates it across the squad.

    Lead
    • GitHub
    • Slack

  2. LinusBackend Engineer

    Designs and implements backend services and APIs.

    Member
    • GitHub

  3. MargaretBackend Engineer

    Implements data models, jobs, and integrations.

    Member
    • GitHub

  4. GraceFrontend Engineer

    Builds the user interface and covers it with tests.

    Member
    • GitHub

  5. TessQA Engineer

    Finds regressions and writes clear reproductions.

    Member
    • GitHub

  6. DennisDevOps Engineer

    Keeps CI/CD, environments, and observability reliable.

    Member
    • GitHub
blueprints
12
One per department, each a complete AI team.
AI agents
52
Each with a name, a role and a remit.
connectors
44
The systems the teams are already set up for.

Approvals

What cannot be undone waits for you.

Every blueprint is written the same way: an AI agent prepares the step and a person decides. A refund, a reply to a customer, a change in your accounts. The platform shows exactly what is about to happen, and nothing runs before you say so.

How approvals work
Approvals · Webshop deskExample

Needs your decision

Refund € 24.95 on order 4182

Creating a refund in Mollie always waits for a person.

Amount
€ 24.95 of € 124.95
Payment
tr_7UhSN1zuXS
Reason
Mug arrived broken
Requested by
Sanne · Payments Agent · Webshop desk

Yours to change

Change what you like.

Give an AI agent a different name, model or rank. Move it to another team, or add one the blueprint did not think of. Nothing in a blueprint is fixed once it is deployed.

None of them fits? Describe your department to Pal and correct what it proposes, or build it by hand. The structure and the controls are the same.

More about Pal

Every department.

Open one to see the team, the work it takes on first and where it waits for you.

The icons are the systems the blueprint gives the team. What each one can change, and what waits for a person, is on the connector’s own page.

On the platform

The same controls under every team.

A blueprint decides who does the work. How that work is governed is the platform’s job, and it is the same for every team.

  • Approvals

    An action that changes something waits for a person, who sees the exact call before it runs.

  • Budgets and limits

    Each department has a cap on tasks at once. Add a monthly budget and new runs stop at the cap.

  • Routing

    A model per AI agent, with up to three fallback models for when a provider is down.

  • Audit trail

    Every run is on the record: who did what, in which system, and what it cost.

Rolling out

One step at a time.

A department gets its freedom step by step. Each step asks that the one before it has become routine, and at every step a person stays in the picture.

How control works

  1. Shadow first. AI agents draft, people send. Nothing that changes something is switched on.

  2. Then with approval. Real tools, and every change is decided by the person you name.

  3. Then partly on their own. You exempt the specific actions that have earned it. Budgets and delegation limits stay.

  4. Eval cases from day one, so a small change to an instruction cannot quietly break what already works.

Start with your busiest queue.

Tell us which one. We listen to your question first, then walk one real piece of work through the team while you watch the approvals.

Book a demo