Loading page
Point your existing client at Tsubasa and choose Fast or Pro. Your prompts, outputs, and source code stay out of usage history.
How do I connect an OpenAI-compatible coding agent to Tsubasa?
Preparing stream…
This animated interface is a product demonstration; it does not send a request.01 / MODELS
Use Fast when speed and volume matter. Use Pro when the task needs deeper reasoning or more complicated code.
Selected model
tsubasa-pro
Our flagship model for coding, reasoning, agents, and demanding professional work.
Ready for API requests in this environment.
02 / CLIENTS & CODING TOOLS
For an OpenAI client, change the key, base URL, and model ID. For a coding tool, use only the interface its current release supports. Tsubasa provides model discovery and Chat Completions—not every OpenAI or Anthropic endpoint.
OpenAI-compatible clients
Create a named key, choose a model ID from the catalog, and point the client at api.tsubasa.sh/v1.
import os
from openai import OpenAI
client = OpenAI(
api_key=os.environ["TSUBASA_API_KEY"],
base_url="https://api.tsubasa.sh/v1",
)
stream = client.chat.completions.create(
model="tsubasa-pro",
messages=[{"role": "user", "content": "Hello"}],
stream=True,
)
for chunk in stream:
print(chunk.choices[0].delta.content or "", end="")Agent interface previews
These illustrations show Codex, Claude Code, OpenClaude, Hermes Agent, and OpenClaw interfaces. Check the integration pages for tested compatibility. Hermes v0.21.5 requires a larger context window than Tsubasa currently serves.
Five coding tool terminal recreations rotate automatically while this carousel is visible. Use the pause, previous, and next buttons, or focus this carousel and use the left and right arrow keys.
Admission acquires a lease and places a worst-case credit hold.
03 / PREPAID BILLING
Add credit before you start. Tsubasa sets enough aside for each request, then charges for the tokens used. Requests can run at the same time without spending the same balance twice.
THE TSUBASA PATH
DATA HANDLING
Tsubasa does not write prompts, completions, or source code to usage history. It keeps token counts, cost, status, and timing so you can see what happened and what you paid.
0
prompt or completion bodies written to usage history by default
FAQ
If it lets you set an OpenAI base URL, API key, and model ID, it may connect directly. Tsubasa supports model discovery and Chat Completions, including streaming. The integrations page shows what we have tested.
The model page shows what is available in the environment you are using. Local demo models are clearly labelled and do not imply production access.
Add credit before you run a model. Tsubasa sets aside enough for each request, then charges only for the tokens it used.
Tsubasa does not write prompt or completion content to usage history. It keeps details such as the request ID, model, token counts, cost, status, and timing.
You keep the output already received. Once the model provider confirms the stop, Tsubasa charges for the work completed and releases the remaining credit. If the stop cannot be confirmed, the reservation stays in place until recovery resolves it.
Yes. Give each app or environment its own key. You can revoke one key without touching the others.
No. Tsubasa adds credit only after the payment provider confirms the payment. Returning to the site is not enough on its own.
Choose a model, connect the tool you already use, and create a key when you are ready to run your first request.