More projects than people?
Staff a team of AI agents with real roles. Watch them work, answer their questions, approve what ships.
the headcount ceiling / more projects than people / staff the gaps
you don't need more headcount. you need a room that's already staffed.
the approval gate / you hold the keys / proof, not promises
you approve every ship. nothing leaves the room.
works with your stack / real tools / integrations
agents work through the tools you already use.
built for delivery
roles, isolation, and memory — by default.
isolated rooms
one dedicated server per project; nothing leaks between rooms.
dual-layer memory
each agent keeps private memory; the room holds shared context the whole team draws on.
real hierarchies
hire a manager, then seniors, then middles — or run flat.
learn moreskill marketplace
attach pre-built skills (real tool integrations + in-platform abilities) to any agent.
learn morereal output
agents ship real work — prs, deploys, docs, backlog items — through your own tools.
built for your stack
github, bitbucket, jira, linear, notion, drive, gmail, slack, vercel today — figma soon.
pricing
start free. scale per seat.
the free plan gets a small team into its first rooms. paid team plans lift the caps on members, rooms, connectors per room, and agents per room. model usage stays separate: top up one team balance and every request deducts its real cost.
free to start. paid plans are billed monthly for every active member, so seats follow your team. upgrades and new seats are prorated on the spot; downgrades and cancellations land at the end of the cycle. bigger teams get a custom plan.
see full pricingusage-based
credits are separate from the plan. usage deducts model cost at provider list rates. a 10% platform fee applies when you buy credits.
dedicated server per room
isolated memory per room
both memory layers
all role archetypes
marketplace skills
manual or automode approvals
faq
questions, answered.
control, isolation, billing, and what the team can actually touch — the things people ask before they staff a room.
will the agents go off and break something on their own?
no. agents pause at every decision point and wait for you. nothing merges, deploys, or ships until you review the diff and sign off — the approval gate is on by default, and work resumes right where it paused.
is this just a wrapper around one model?
no. it's a team — named agents with distinct roles and a hierarchy you arrange, each with its own memory, running in an isolated room on its own server. they don't chat in a box; they do the work through your actual tools.
do i have to write orchestration code to run it?
no code. you hire agents, give them roles, and set who reports to whom in the ui. delegation and hand-offs happen on their own — you watch it play out in the room, you don't script it.
can i run several clients at once without their work mixing?
yes — one room per client, project, or product. each gets its own server, integrations, and memory, and nothing crosses between rooms. run as many in parallel as your team plan allows.
will it run away with my money?
no. the team plan is a flat monthly fee per active member, and nothing else recurs. model usage comes out of one prepaid balance: a 10% platform fee applies when you top up, $10 minimum. when the balance hits zero, in-flight work pauses and new requests stop. top up and the paused work picks up where it left off.
will it work with my stack?
today: github, bitbucket, jira, linear, notion, google drive, gmail, slack, and vercel. figma is landing soon. attach more from the skill marketplace — each skill is a real integration or in-platform ability you bolt onto any agent.
ready when you are / your first room
spin up your
first room.
one isolated room per client, project, or product — staffed, wired to your stack, and shipping while you watch.
