PRICING
Pay for the work, not the seats.
Humans are free. You pay for staffed spaces and the thinking each task actually needs — every run logged, every token visible.
Pilot
YOUR FIRST AGENT, ON US
One space and one working agent, set up with you on a call. See real work land before you pay anything.
- 1 space · 1 agent
- 2 routines on a schedule
- Core connectors (Gmail, Slack, Calendar)
- Always / Ask / Never permissions
- Model usage included during the pilot
Team
A STAFFED SPACE FOR YOUR TEAM
Every agent, routine, and connector your space needs, with usage billed transparently — you pay for the thinking each task actually needs, at cost.
- Unlimited agents & routines per space
- All connectors + code-mode SDK
- Skills: teach once, reuse everywhere
- Multi-model routing (Claude, GPT, GLM, DeepSeek)
- Isolated sandbox per agent
- Run logs, analytics & spend visibility
- iOS app access
Enterprise
FLEETS, ORGS, AND COMPLIANCE
For operators running many spaces or deploying agents to clients: org-level controls, custom deployment targets, and a signed DPA.
- Multiple spaces under one org
- Self-hosted or dedicated nodes
- SSO & role-based access
- DPA + security review
- Priority support with a named contact
HOW USAGE WORKS
A director model plans. Cheaper models execute. You see the bill per run.
Harby routes each piece of work to the smallest brain that can do it well, so a morning inbox triage doesn't cost frontier-model prices. Analytics shows what each agent did, what it changed, and what it cost — down to the token.
QUESTIONS
Fair questions, straight answers.
How does model usage billing work?
Agents use the model that fits each job — a big reasoning model to direct, smaller ones to execute. You see every run's token usage and cost in Analytics, and usage is passed through transparently rather than marked up behind a bundle.
What counts as a space?
A space is a room for one project or team — its own agents, files, connectors, and routines. Most teams start with one; operators running several clients use one space per client.
Can I bring my own model subscription?
Yes. Teams on Claude or ChatGPT subscription plans can connect them, and Harby routes those agents through your existing plan instead of metered API usage.
What happens when the pilot ends?
Nothing abrupt. Your agents keep their files, memory, and routines — we simply move the space onto Team pricing you agreed on the call, or wind it down and export your data if it isn't a fit.
Do you charge per seat?
No. Humans are free — invite your whole team into a space. You pay for the agents' work, not for watching it.