Run the harness yourself, or let us run it.
bookistudios AI is MIT-licensed and free forever when you self-host. The hosted plans add managed sandboxes, persistent memory, and scheduled agents — at roughly a third of what hosted-only agent products charge for the same capability tier.
Self-hosted
Your machine, your keys
MIT license · no account needed
Read the install guide- Full agent harness, unlimited runs
- Bring your own model keys
- Local, Docker or Kubernetes sandboxes
- All built-in and custom skills
- Sub-agents, memory, MCP servers
- Community support on GitHub
Free
Hosted, no card
No credit card required
Start free- 300 agent runs / month
- 60 messages / day
- 2M tokens / month
- 1 concurrent task
- 5 scheduled tasks · 2 projects
- 1 GB workspace storage
Pro
For daily driving an agent
$228 billed yearly
Subscribe- 4,000 agent runs / month
- Unlimited messages
- 25M tokens / month
- 4 concurrent tasks · 4 sub-agents
- Unlimited scheduled tasks · 50 projects
- 50 GB storage · long-term memory
- Deploy sites and dashboards from a run
Team
Shared skills and runs
$588 / user billed yearly
Start a team- 10,000 agent runs / user / month
- 60M tokens / month, pooled
- 8 concurrent tasks · 8 sub-agents
- Shared skill library and agents
- Roles, seats and usage dashboard
- 200 GB pooled storage
Enterprise
Self-host with a contract
Annual, invoiced
Talk to us- Unlimited runs and seats
- VPC or on-prem deployment
- SSO / SAML, audit logs, IP allowlist
- Private model routing
- SLA, dedicated support channel
- Security review and DPA
Hosted plans include model inference on our managed keys. Bring your own keys on any plan and runs stop counting against your token allowance.
Compare plans
| Capability | Self-hosted | Free | Pro | Team | Enterprise |
|---|---|---|---|---|---|
| Agent runs / month | Unlimited | 300 | 4,000 | 10,000 / user | Unlimited |
| Messages / day | Unlimited | 60 | Unlimited | Unlimited | Unlimited |
| Included tokens / month | Your keys | 2M | 25M | 60M pooled | Negotiated |
| Concurrent tasks | Hardware-bound | 1 | 4 | 8 | Custom |
| Sub-agents per run | Unlimited | 1 | 4 | 8 | Custom |
| Scheduled tasks | Unlimited | 5 | Unlimited | Unlimited | Unlimited |
| Workspace storage | Your disk | 1 GB | 50 GB | 200 GB pooled | Custom |
| Long-term memory | Local | 7-day | Persistent | Persistent + shared | Persistent + shared |
| Custom skills & MCP servers | Yes | 3 skills | Unlimited | Unlimited + shared | Unlimited + shared |
| IM channels (Slack, Telegram…) | Yes | 1 channel | All channels | All channels | All channels |
| Deploy sites with a database | Self-managed | — | Yes | Yes | Yes |
| Support | Community | Community | Email, 2 business days | Priority email | SLA + shared channel |
| SSO / SAML, audit logs | — | — | — | SSO | SSO + SAML + audit |
Questions
What counts as an agent run?
One task the lead agent accepts and carries to completion, including every sub-agent it spawns. Follow-up messages in the same thread are not new runs.
What happens when I hit a limit?
Runs queue instead of failing. You can add your own model keys to keep going without a token allowance, or upgrade mid-cycle and only pay the prorated difference.
Is the self-hosted version limited?
No. It is the same MIT-licensed harness under the same repository — skills, sub-agents, sandboxes, memory, IM channels. Hosted plans pay for infrastructure and managed inference, not features.
Can I move between self-hosted and cloud?
Yes. Threads, memory and skills export as plain files and import into either. Your config.yaml carries over unchanged.
Refunds?
Cancel any time from the billing page. Annual plans are refundable in full within 14 days.