Your agents deserve an open tab.
OpenAI-compatible inference with published rates. Pay as you go or subscribe for up to 8× usage. Your prompts are never stored.
Request an Invite
- OpenAI-compatible
- Published rates
- Hosted in the USA
- Zero data retention
How it works
Get your API key
Request an invite, choose how to pay, and create a key.
Point your client
Change one base URL. Keep your OpenAI-compatible client or tool.
client = OpenAI(
base_url="https://api.cirthan.com/v1",
api_key="sk-..."
)
Build
Run chat completions or transcription at published rates.
Models & rates
Two inference models with clear input and output rates.
Qwen3.6 35B A3B
Qwen3.6 27B
Text rates are per million tokens; transcription is per hour of audio.
What you can build
Always-on assistants
Power OpenClaw, Hermes Agent, and other always-on assistants through one OpenAI-compatible API.
Audio pipelines
Transcribe meetings, podcasts, and voice notes, then route the text into your agents.
Batch inference
Classify, extract, and summarize at scale with predictable published rates.
What makes Cirthan different
Choose how you pay
Pay as you go, or subscribe for up to 8× usage.
Pay only for what you use
Buy credits, use any model, and switch anytime.
- Usage at published rates
- Top up anytime
- Credits carry forward
Get up to 8× usage
Use any model. Switch anytime. Usage resets monthly.
- 8× usage on Qwen3.6 35B A3B
- 5× usage on Qwen3.6 27B
- Switch models anytime
Subscription usage resets monthly. Purchased credits carry forward.
Cirthan is invite-only
We onboard developers in small batches. Leave your email and we'll send an invite when a slot opens.
No spam. No mailing list. Just an invite.
Already have an invite code? Sign up now