Pay only for what you use. No seats, no monthly minimums, no surprises. Billed per 1 million tokens.
Grounded retrieval, symbolic planning and memory that persists between sessions, around a swappable LLM. Answers by default, works with Claude Code, Cline, Cursor and Windsurf, and bills on the tokens it actually uses.
$0.14
per 1M tokens (in + out)
Prefix a model with raw/ to skip the engine and go straight to that provider — same key, same endpoint.
| Model | Provider | Input | Output | Context |
|---|---|---|---|---|
GLM-5.2Best for coding | Z.ai | $0.14 | $0.14 | 1M |
GLM-5.1Long-horizon | Z.ai | $0.14 | $0.14 | 200K |
GLM-5Flagship open-source | Z.ai | $1.00 | $1.00 | 200K |
GLM-4.5Ultra cheap | Z.ai | $0.07 | $0.07 | 128K |
GPT-4o | OpenAI | $5.00 | $15.00 | 128K |
GPT-4o mini | OpenAI | $0.15 | $0.60 | 128K |
Claude 3.5 Sonnet | Anthropic | $3.00 | $15.00 | 200K |
Claude 3.5 Haiku | Anthropic | $0.80 | $4.00 | 200K |
* Prices are per 1,000,000 tokens (input or output)
1M code completions (avg 500 tokens)
100K agentic coding sessions (avg 5K tokens)
10K complex refactors (avg 8K tokens)
Billing starts after you add a payment method in the dashboard. You can test for free without a card.
We count prompt tokens (input) and completion tokens (output) separately. Usage is logged per request and visible in your dashboard in real time.
Yes — set a monthly cap in your dashboard. Requests will return a 429 once your limit is reached.