Token API Guide

1. Create an account

Register for a TATC account and sign in to the Token API console. The console is the central place for API keys, model access, usage and billing information.

2. Open API Key management

From the console, open API Key management and choose to create a key for an application, environment or team workflow.

3. Configure the key

Console fields and controls are defined for the selected service arrangement.

• Give the key a clear name for its application and environment.
• Select the upstream models the key may access.
• Configure rate limits and usage quotas for controlled spending.

4. Store and manage the key safely

• Store the key in a server-side environment variable or secret manager.
• Keep it out of browser code, mobile apps, public repositories and screenshots.
• Disable or rotate a key when an application changes or exposure is suspected.

1. Select an upstream model

Browse the catalogue, compare model capabilities and choose the upstream model route that fits the workload. Copy the model ID shown by the platform.

2. Prepare your connection values

Set the gateway endpoint, API key and model ID as server-side environment variables. The names below illustrate the integration pattern.

• TATC_BASE_URL: confirmed API base URL
• TATC_API_KEY: issued access credential
• TATC_MODEL_ID: model identifier enabled for your arrangement

3. Send a small test request

The cURL example shows an OpenAI-compatible request pattern. Replace each placeholder with values shown in the console.

# Example integration pattern. Use the details supplied by TATC.
export TATC_API_KEY="<issued-key>"
export TATC_BASE_URL="https://<confirmed-host>/v1"
export TATC_MODEL_ID="<confirmed-model-id>"

curl "$TATC_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $TATC_API_KEY" \
  -H "Content-Type: application/json" \
  -d "$(printf '{"model":"%s","messages":[{"role":"user","content":"Hello"}]}' "$TATC_MODEL_ID")"

4. Use a compatible client

The gateway supports familiar OpenAI- and Anthropic-style integrations. Configure a supported client with the gateway base URL, API key and selected model ID.

# Example OpenAI-compatible pattern.
import os
from openai import OpenAI

client = OpenAI(
    base_url=os.environ["TATC_BASE_URL"],
    api_key=os.environ["TATC_API_KEY"],
)
response = client.chat.completions.create(
    model=os.environ["TATC_MODEL_ID"],
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

5. Check the response and usage

Streaming, tool calls and multimodal inputs depend on the selected upstream model. Add them after the basic request works.

• Confirm the HTTP result and returned content.
• Record a request identifier when one is returned.
• Open the console to review model, API-key, token and billing records.

Model and protocol selection

Choose a supported upstream model and use the compatible route shown in the catalogue. OpenAI- and Anthropic-style interfaces let existing SDKs and developer tools connect by changing the base URL, key and model ID.

Authentication and key changes

Use each key only from trusted server-side systems. Document which application uses it and use the console controls to disable or rotate a key when access changes or exposure is suspected.

Handling request failures

These are common integration patterns. Refer to the selected service documentation for service-specific status codes, error bodies and retry timing.

• An authentication or permission response calls for checking the key, endpoint and model access before retrying.
• If a limit response is returned, follow the agreed retry guidance and avoid rapid repeated requests.
• For a transient service error, retain the time, request identifier if present and a redacted request summary for support.

Usage and billing

The console brings together request counts, token usage, model-level and API-key-level consumption, quotas and billing information. Pricing and billing presentation depend on the selected service arrangement.

When contacting support

• Include the affected service and model ID, approximate time and a request identifier if one was returned.
• Describe the observed status and whether the issue is repeatable.
• Remove API keys, personal data and confidential prompts from screenshots and logs before sharing them.