Agents – your autonomous team

Agents are AI employees. Where Chat waits for your brief and Journeys run the automation you designed, an agent owns a piece of your marketing and decides its own next move – it proposes the work, does it, and brings you the results to approve. You stay in charge of what goes live.

Agents are in beta and available on the Manager and CMO plans. The workspace lives under Agents in the sidebar, with three tabs: Today (the work waiting for you), Team (your roster), and Office (the rules they work under).

Two ways to hire

There are two kinds of agent, and the difference is the same as in a real team:

• A specialist is an employee. It owns one piece of your marketing permanently – your ads, your search visibility, your social – and keeps proposing and doing work in that area for as long as it's on the team.
• A one-off goal is a contractor. You describe a single outcome you want, the agent researches your business, plans the checkpoints, and works until it's done – then it's finished. No standing role, no ongoing cadence.

Both produce drafts into the same Today review queue, obey the same trust levels and spending limits, and show up side by side under your active goals. Hire an employee when an area should never go quiet; hire a contractor when you just need a specific job done.

Hiring a specialist

The Team tab offers eight specialists, each built to own one area:

Every card states what the specialist does, what it needs (connected accounts, some history to learn from), and what it costs to think. You can rename and rebrief any of them, or build a fully custom specialist from scratch.

After hiring, the specialist reviews your setup and asks whatever it still needs – one question per screen, then Save and let them start.

Hiring a one-off goal

For a single job, choose hire a one-off goal on the Team tab and describe what you want done in plain words. The agent researches your business, plans the checkpoints, and starts working – you can edit or pause it at any time. Its drafts arrive in Today like everyone else's, its goal sits in the active-goals list alongside your specialists' work, and when the outcome is reached, the job is done.

How they work

On their own cadence – roughly weekly, or on demand with Bring me fresh ideas – each specialist proposes a handful of goals from your data, each viewed through a different lens (compounding wins, new capabilities, research, creation, optimization). What happens next depends on the trust you've given them:

• Asks me before publishing – ideas and drafts queue for you; nothing ships until you approve.
• Waits 24h, then publishes – drafts auto-publish a day after creation unless you reject them first.
• Publishes immediately – work goes live as it's produced; you're notified after.

Trust is set globally in the Office and can be overridden per specialist. Everything they produce lands in the Today tab as drafts – approve or reject each one, and filter by what needs review, what's approved, published, or rejected.

The Office: rails they can't cross

• Master switch – pause the whole autonomous operation for a brand with one toggle.
• Spending limits – credits per day, credits per week, and drafts per day. When a limit is reached, the team stops until tomorrow. These limits never touch your own manual usage.
• Working hours – active days, quiet hours, and timezone. Nothing publishes at 3am unless you want it to.
• Credential vault – saved sign-ins (encrypted, 2FA supported) for the platforms your agents work in, used only when an agent needs to log in for you.
• Trigger rules – "when a condition is met, your team does something automatically."

The Social Media Manager learns your channels

The social specialist keeps a playbook per platform – what wins on LinkedIn often flops on X, so rules never cross platforms. You rate each published post (Worked / Meh / Flop); once a platform has three rated posts, the weekly reflection proposes rules – what works, what doesn't, voice, timing, formats – and you accept or reject each one. Accepted rules shape everything it makes next. You also choose its mode: original posts, joining conversations, or both.

Replying and engaging happens through your saved logins or the built-in cloud browser session – so it works on platforms without a posting API, and every session is visible in the activity log.

Watching them work

Each specialist's profile carries an Activity timeline – goals started and completed, drafts published (with links to the live posts where the platform provides them), your verdicts, and playbook rules learned. The roster shows at a glance who's working, who's paused, and who's waiting on an answer from you. Pause, resume, or let a specialist go at any time.

What it costs

A batch of fresh ideas costs about 2 credits. Executing goals bills per tool used (1–50 credits by complexity, plus a small per-action charge), and a cloud-browser session – used for platforms without a native API – costs 40 credits. Your daily and weekly limits in the Office are the hard ceiling; agents stop before they pass them.

Agents, Journeys, or Chat?

• Chat – you drive, one brief at a time.
• Journeys – you design the automation; it runs exactly that, forever.
• Agents – you set the goal and the rails; they decide the moves.

They compose: an agent can build journeys, and everything any of them produce respects the same suppression, unsubscribe, and frequency rules as work you ship yourself.