Pay for the platform. Your model tokens stay at cost.
Two layers, nothing hidden: a flat platform fee for the runtime, Studio and governance — plus usage billed per agent run. Bring your own model key, or let us manage it at provider cost. Run it in our cloud or your own infrastructure.
For builders kicking the tires.
- 1 agent · 1 seat
- Haiku model
- Thaliq Studio + API & SDK
- In-app web widget
- Conversation logs
- Shared cloud · community support
For startups shipping their first agents.
- 5 agents · 3 seats
- All models · multi-model routing
- HITL gates + guardrails
- Tools / MCP gateway (3 servers)
- WhatsApp + Telegram channels
- Traces (OTEL) + spend alerts
- Email support · prepaid wallet
For teams scaling agents across customers.
- 25 agents · 10 seats
- Model fallback · BYOK optional
- Sub-agents · unlimited MCP
- Isolated RAG + re-ranking
- SSO add-on · RBAC · token passthrough
- Shared or dedicated · multi-region
- Priority support · net-30 billing
For regulated and high-isolation deployments.
- Unlimited agents · custom seats
- Dedicated or self-hosted (commercial license)
- SSO / SAML included · Zero Data Retention
- HIPAA / BAA · audit unlimited + export
- PrivateLink / VPC peering · BYOK secrets
- White-label · custom domain
- Dedicated CSM · 99.97% SLA
Prices in USD. Annual billing saves ~17% (two months). Usage is billed separately, per run.
Two layers. No surprises.
A flat platform fee for the product, and usage for what your agents actually run. That is the whole model.
A predictable monthly or annual fee for the runtime, Studio, governance and the support of your tier. This is the product you pay for.
Billed per agent run. On managed billing you pay model tokens at provider cost; on BYOK you bring your own key and we never touch your model spend.
Self-serve plans run on a prepaid wallet with auto-reload, so an agent never stops mid-task. Business and Enterprise can move to net-30 or an annual commit.
We do not mark up your tokens. The platform is what you pay for — your model spend goes to the provider at cost, whether we manage the key or you bring your own.
Everything, side by side.
The full picture. Usage is billed per run on top of every paid plan; the platform fee is what changes by tier.
| Free | Pro | Business | Enterprise | |
|---|---|---|---|---|
| Platform & access | ||||
| Thaliq Studio | Included | Included | Included | Included |
| API + SDK | Included | Included | Included | Included |
| CLI + versioned spec | Not included | Included | Included | Included |
| Environments | Not included | dev + prod | full | full + custom |
| Active agents | 1 | 5 | 25 | Unlimited |
| Seats | 1 | 3 | 10 | Custom |
| Models & orchestration | ||||
| Multi-model routing | Not included | Included | Included | Included |
| Model selection | Haiku | All | All | All + self-host |
| Automatic model fallback | Not included | Not included | Included | Included |
| Prompt caching | Included | Included | Included | Included |
| BYOK (bring your own key) | Not included | Not included | Optional | Included |
| Agents & capabilities | ||||
| Sub-agents / composition | Not included | Not included | Included | Included |
| HITL gates | Not included | Included | Included | Included |
| Guardrails | Basic | Included | Included | Advanced |
| Tools / MCP gateway | Not included | 3 servers | Unlimited | Unlimited |
| Memory (short / long term) | Short | Short + long | Full · pgvector | Full · dedicated |
| Channels | ||||
| In-app web widget | Included | Included | Included | Included |
| REST API + SDK | Included | Included | Included | Included |
| WhatsApp + Telegram | Not included | Included | Included | Included |
| Security & isolation | ||||
| Tenant isolation | Shared | Row-level | Row-level | Schema / DB dedicated |
| SSO / SAML | Not included | Not included | Add-on | Included |
| RBAC | Not included | Basic | Included | Included |
| Token passthrough | Not included | Not included | Included | Included |
| Audit logs | Not included | 30 days | 1 year | Unlimited + export |
| HIPAA / BAA | Not included | Not included | Not included | Included |
| Deployment & support | ||||
| Deployment mode | Shared | Shared | Shared or dedicated | Dedicated or self-hosted |
| Regions | us-east-1 | us-east-1 | Multi-region | Custom |
| Support | Community | Email · 48h | Priority · 8h | CSM · 1h |
| Uptime SLA | Not included | 99.5% | 99.9% | 99.97% |
Before you pick a plan.
What counts as a run?
One agent execution — a single turn that may fan out into several model calls and tool calls. We bill per run; the model token cost is separate and billed at provider cost.
Do you mark up model tokens?
No. On managed billing you pay tokens at the provider’s published price; on BYOK you bring your own model key and we never see your spend. We make our money on the platform, not your tokens.
What is BYOK?
Bring Your Own Key — route inference to your own Anthropic, Bedrock or OpenAI account, so the model bill and the key stay on your side. Available on Business (optional) and Enterprise.
Can I self-host the runtime?
Yes, on Enterprise, under a commercial license — in your VPC or on-premise. Your data never leaves your network, and you run the exact same runtime as our cloud.
What about compliance?
GDPR with a DPA is available today. SOC 2 Type II and ISO 27001 are in progress. HIPAA with a signed BAA is available on Enterprise. We keep the security page honest about current status.
Can I change plans later?
Yes — upgrade or downgrade at any time. The platform fee prorates and usage keeps billing per run, so nothing breaks mid-cycle.
Start free.
Move up when it matters.
500 runs a month, no card. Per-tenant isolation, audit on, plan limits enforced. Upgrade for SLA and white-label, or self-host the runtime when control matters.