Cohort
Agentic AI workflows, designed and operated

Design agent workflows easily.
Know what every run costs.

Start from a ready-made workflow or describe your own. Set a spending limit, then watch each run as it happens: what it is doing, and what it is costing you.

Design a workflow See how it works
Running real products today.
Who runs on it

Already doing real work.

Real products, real traffic, on the same limits and receipts you get.

live OmniDeck

Assistants people build by talking to one.

omnideck.ai live Trellis

Engineering agents that ship under governance.

xtrellis.ai in pilot Hakeem

Patient intake, in the patient's own Arabic.

hakeem.projexlabs.com live Our own fleet

The agents that build and run our products.

projexlabs.com
What it solves

Agents are easy to demo. Hard to run as a business.

Every new workflow means wiring prompts, tools and guardrails by hand.

Design workflows, not plumbing

Start from a ready-made workflow and make it yours: what it must find out, what it must never do, and when it is done. Cohort runs it.

Must find out · must never · done when

You find out what your agents cost when the invoice arrives.

See the unit economics

Cost per run, per workflow and per customer, while it runs. Set a spending limit before a run starts, and it stops there.

Cost against the limit, settled or not

When an agent goes wrong, all you have is a transcript to dig through.

Watch every run as it works

See each rule being met, what the run is waiting on, and where it went off-script. Get a receipt when it is done.

On the run page: the rules, live
How it works

Three steps from idea to receipt.

01

Pick a workflow

Start from a ready-made one, or describe your own in plain language.

02

Set a limit

Decide what a run may spend before it starts. It stops there.

03

Run it and watch

Follow it live, then get a receipt: what it did, which rules it met, what it cost.

Where your money sits

Credit moves down in three steps, and only when you move it: your account, then a project, then an access key (what your software uses to start runs). A run spends only what its key was given, so an agent can never spend more than was put in front of it. Only a signed-in person can move credit into a project.

How a run proves what it did

Give a workflow rules (what it must find out, and when it is done) and Cohort records the run against them as it works. At the end there is a record of which rules were met, built as it went rather than written afterwards by the agent. It is optional: reach for it when the work may need defending later.

What it costs

Every run is measured and itemised, and its total says whether it has settled yet. Prices depend on the model, and are published live in the console under Price list rather than quoted here, where they would go stale.

Design your first workflow.

Pick one, set a limit, and watch it run.

Sign in See how it works Who runs on it