Teams of agents

Put several personas in one room with a code executor: a selector that lets a decision model pick who speaks next, a swarm or an orchestrated crew, runnable as a step in any routine.

Some work needs more than one voice. A report that has to be researched, computed and written is three jobs with three temperaments, and a single persona doing all of them does each a little worse.

One team's page: its month of runs, the model it runs on, how it takes turns, and the six agents on the roster.

Roles in a room

A team is a roster of personas with a way of taking turns. A selector team chooses who speaks next from what has been said so far - the everyday choice, and the fastest. Hand-off rules settle the turns that are not a judgment; the decision model settles the rest in half a second, from a one-line description of when each agent speaks, and falls back to the lead below a confidence you set. A swarm hands the floor directly from agent to agent, for work where the handoff is the point. An orchestrated crew plans, delegates and checks the plan against progress, for longer jobs that need a manager. Each team can carry a code executor: a sandboxed container where the team's coder runs what it writes, so numbers in the report were computed rather than remembered.

Watch it work

Teams run in a live chat. You see each turn as it happens, which tool ran and what came back, and when the team declares itself done. Execution metrics - turns, tokens, time, tool calls - are recorded per run, so a team that is slow or expensive can be seen and reshaped rather than guessed at.

A step in a routine

A team can be a step in any routine, next to a persona step or a publish step. A digest routine can hand its gathered material to a team of analysts and writers, take the finished report back, and publish it - on a schedule, without anyone in the loop. Mixed models are fine: the writer can run on a different endpoint from the analyst if that is where the strength is.

On your keys, in your workspace

Teams run on your organization's provider credentials and their files land in a workspace you can open, browse and download from. The executor container belongs to the run and is removed when the run ends; the files stay.

All features Talk to us

More of what it does

Jarvis, the expert at your fingertips

Ask Jarvis in plain words and it does the work in every module on your plan: reschedule a routine, copy a team, draft an article. It asks before anything changes, naming exactly what.

Jarvis remembers, learns and follows through

Jarvis remembers you, your goals and your decisions, learns after each conversation, and lists every change for you to undo. It keeps how you like things done and checks in when something waits on you.

Jarvis shows its work, and you act on it

While it talks, Jarvis puts what it found on screen: figures, charts, tables, maps, video, your day. Many cards take an answer too: tap a choice, tick a checklist, fill in a form.

Jarvis, where your people already are

Reach Jarvis on Slack, Teams, WhatsApp, Telegram, a phone or email, and run your home and workflows through it, each on your own account with limits you set. Webhooks work both ways.

Missions: long work, with a team and a swarm

Give Jarvis a goal too big for one answer and it works it in the background: a plan, a budget, a team and a swarm, and an independent check before it calls the work done.

The whole platform, on Jarvis's desk

Ask for the thing itself and Jarvis puts its page on the desk: an article, a chapter, a knowledge graph turning in 3D, a dashboard. Each card comes back as you left it.