The product’s words, defined

Plain definitions for the words the product and this site use. Each links to where it’s explained in full.

Agent: An AI worker with one job, a brief and a funnel; some also have a budget. It works every day, proposes changes, files reports, and is managed like a member of staff.
Brief: What an agent should achieve, who it should reach and leave alone, what it must never do, its channels, and the target it’s judged on.
Claim: A figure or promise agents may use, with its source. A claim without a source can’t run.
Conversion event: Something people do that counts as a result, such as an order or a qualified lead, sent from your store, your CRM or the events API, with what it was worth.
Decision pace: How often an agent makes material changes, set by how fast its funnel produces enough evidence.
Desk: An agent’s area of work. Some desks buy media or send email; brand review, the creative desk and the copy desk serve the others.
Executor: The one part that changes your accounts, and the only part that holds their access. It runs the rules again before a change you approved, makes the change and records the result.
Experiment: A test of one idea, with a hypothesis, a split and a planned length. The software calculates confidence.
Funnel: The ordered steps people take toward a result, such as added to cart, checkout started and first order. Each agent has its own.
Guardrail: A limit no agent can cross whatever it proposes and whatever you approve: spend caps, contact limits, consent, content rules.
Guidance: A note you give an agent. It goes in the notebook and keeps applying; to change it, give new guidance.
Holdout: People deliberately left out of a send, so its effect can be measured against people who didn’t get it.
Inbox: Where proposals wait for you (To decide), and where you read what agents did (To read, Silent, History).
Library: The company’s approved images and copy. Agents can use only approved items.
Notebook: An agent’s memory: hypotheses, experiments, decisions, learnings and your guidance, each with its status.
Permission level: Who decides an action, one of four levels: Not allowed, You approve, Acts and notifies you, or Acts silently. Set per action and per agent.
Playbook: Defaults for a kind of business (store, lead generation, subscription, marketplace): a starting funnel, the audiences to leave out, and a target.
Policy engine: The software that checks every proposal against your permission levels and guardrails, then lets it through, stops it or sends it to you. No model is involved.
Proposal: A change an agent wants to make, with what it changes, why, and the checks it passed. Agents can only propose.
Record: The history of every action: who proposed it, why, the checks, who approved it and what happened.
Report: What an agent files on its cadence: what happened, why, what it did and what it plans next.
Stop button: Pauses every agent at once. Campaigns hold their last settings.
Too early to call: What a report says when the evidence isn’t in yet, instead of guessing.
Worth: What one step of a funnel is worth to an agent: a fixed amount, or a value carried by the event, such as the order value.

Give your first agent a brief.

Your work email is all we need to start.

Or try the live demo first