Skip to content
UseAgent

PRICING

UseAgent pricing.
Free while in early access.

Early access is open: leave your email and we let you in, no card. You bring your own provider keys, so model spend bills to you. Self-hosting the open-source platform is free. Enterprise is a conversation.

CLOUD

DEFAULT

Free while in early access

Hosted by UseAgent. Early access is open; leave your email and we let you in.

Join early access

Your provider and computer spend bills to your own accounts.

  • Five engines: Chat, OpenCode, Claude Code, Codex and Pi
  • Bots, routines, handoffs and child sessions
  • Cube as the hosted computer by default. Connect your own Daytona, Cube or Box account and your threads run there
  • Approval gates with a durable audit trail
  • Artifacts, knowledge, skills and team memory
  • Slack and GitHub

SELF-HOST

Free, open source

The public repository on your own machines. One command brings up the stack.

Read the install guide

  • Full platform source at github.com/useagenthq/useagent
  • One command brings up backend, frontend and sandbox
  • Your provider keys, your computers, your database
  • Install guide and reference docs at useagent.org/docs
  • Questions and bugs through GitHub issues

ENTERPRISE

Talk to us

Self-hosted on your infrastructure or a dedicated deployment, with the controls your org needs.

Email us

  • Self-hosted on your infrastructure, or dedicated
  • Org controls, SSO and audit requirements worked out with us
  • Approval policies
  • Audit exports
  • Support from the people who built it

COMPARE

What each plan includes

Engines
CLOUDChat, OpenCode, Claude Code, Codex, Pi
SELF-HOSTThe same five, on your keys
ENTERPRISEThe same five, on your keys
Bots and routines
CLOUDIncluded
SELF-HOSTLands with v0.0.4. The public repo is at v0.0.3 today
ENTERPRISEIncluded
Computers
CLOUDCube by default. Connect your own Daytona, Cube or Box account
SELF-HOSTDaytona or Cube on your account. Box lands with v0.0.4
ENTERPRISEYour provider, or a dedicated pool
Approvals and audit
CLOUDApproval gates, durable event log per run
SELF-HOSTThe same, in your database
ENTERPRISEApproval policies and audit exports
Knowledge and memory
CLOUDKnowledge, wiki, skills, team memory
SELF-HOSTThe same, in your database. Team memory is optional
ENTERPRISEThe same, on your infrastructure
Slack and GitHub
CLOUDIncluded
SELF-HOSTBring your own Slack app and GitHub app
ENTERPRISEYour org's apps, set up with us
Hosting
CLOUDUseAgent Cloud
SELF-HOSTYour machines
ENTERPRISEYour infrastructure or a dedicated deployment
Support
CLOUDEmail and GitHub issues
SELF-HOSTGitHub issues
ENTERPRISEDirect support, scoped with us

Questions, answered.

Do I need my own API keys?

Yes. UseAgent talks to the model through your key, so the spend shows up on your OpenAI, Anthropic or other provider account with no markup from us. Add the key once and every engine uses it.

What does a computer cost?

A computer is the sandbox an engine runs in. Cloud starts you on Cube, the hosted default. Connect your own Daytona, Cube or Box account and the sandbox time bills to that account at the provider's own rates. We add nothing on top and we do not set those rates, which is why there is no computer price on this page.

Can I switch from cloud to self-host and keep my data?

Cloud and self-host are the same platform with the same database schema, so nothing about your runs, threads, knowledge or skills is tied to our hosting. There is no one-click export yet. Email hi@useagent.org before you move and we will work out the transfer with you.

What happens when early access ends?

We will publish prices on this page before we charge anyone, and we will email you ahead of time. Self-host stays free and open source regardless. Your provider and computer bills do not change because they were never ours.

Is there a per-seat price?

Not today. There are no published price points yet, per seat or otherwise. Invite your team to your org and everyone works from the same knowledge, skills and memory. When pricing arrives, it will be here first.

What counts as a run?

A run is one turn of work. You send a message, a routine fires, or a bot is handed a task. An engine picks it up, works in its sandbox, and finishes with a result and a full event log. A thread is a series of runs that share context. We do not meter or bill runs today. What you pay for is the model tokens and sandbox time behind them, to your own providers.

Start with one task.
See what you can hand off.

Explore a completed run, or join the early-access list to hear about getting your team started.