Plans & Limits Reference
The full, canonical limits table for Free, Pro, and Business plans.
Plan limits
| Free | Pro | Business | |
|---|---|---|---|
| Price | $0/month | $19/month | $99/month |
| Monthly messages | 50 | 2,000 | 10,000 |
| Agents | 1 | 3 | 10 |
| Max input tokens / message | 4,000 | 8,000 | 16,000 |
| Max output tokens / message | 1,000 | 2,000 | 4,000 |
| Training data cap | 20 MB | 50 MB | 200 MB |
| Pages per crawl | 25 | 100 | 250 |
| Pages per agent | 50 | 500 | 2,000 |
| Weekly website auto-retrain | No | Yes | Yes |
| Overage | Not available | $0.01 / extra message | $0.008 / extra message |
Message and token limits can be overridden platform-wide by an administrator (see Plan Limits & Overrides); the table above reflects the built-in defaults.
Model access
Every plan currently allows the same catalog of models (GPT-3.5 Turbo, GPT-4 Turbo, GPT-4, Claude 3 Haiku, Claude 3 Sonnet, Claude 3 Opus, Llama 3 70B, Mistral Large), but each plan has a different default model:
| Plan | Default model |
|---|---|
| Free | GPT-3.5 Turbo |
| Pro | GPT-4 Turbo |
| Business | Claude 3 Sonnet |
An agent's model list can also be driven by database-configured models (see Model Configuration), which take precedence over the hardcoded defaults above when present.
Overage billing
Pro and Business plans allow sending messages past the monthly limit instead of blocking the agent; those extra messages are billed per-message as shown above. The Free plan has no overage — agents on Free stop responding once the monthly limit is reached until the next billing cycle.
Per-agent plan overrides
A platform administrator can assign a specific plan tier to an individual agent, which makes that agent use the assigned plan's limits and model access instead of its workspace's plan. This is an admin-only capability — see Plan Limits & Overrides.