Hosted API reference

OpenAI-compatible routing with TEKIZ.AI accounts.

Use one hosted base URL, generated sk_tekiz_* keys, route profiles such as tekizai/frontier, and public model discovery without opening anonymous model traffic.

Create API key Finish setup key One-subscription review Quickstart Gateway migration Troubleshooting Model catalog Pricing

Create an account now. Connect an authorized provider or add billing when you are ready to route.

Hosted API base URL: https://api.tekiz.ai/v1. Browser account and billing UI live at https://app.tekiz.ai.

Auth

Bearer key

Send Authorization: Bearer sk_tekiz_.... Generated keys are scoped to a TEKIZ.AI account and should be rotated from the account page if exposed.

OpenAI

Drop-in clients

Point OpenAI-compatible SDKs and tools at https://api.tekiz.ai/v1. Use tekizai/frontier as the first route profile, or tekizai/fusion on Pro/Max for judged multi-model synthesis.

Limits

Quota aware

Plans enforce monthly request quotas and request-per-minute limits. Responses include safe quota and rate-limit headers for client backoff.

Client setup

Use OpenAI SDK configuration.

export OPENAI_BASE_URL=https://api.tekiz.ai/v1
export OPENAI_API_KEY=sk_tekiz_your_key_here

curl "$OPENAI_BASE_URL/chat/completions" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "User-Agent: OpenAI-compatible-client/1.0" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tekizai/frontier",
    "messages": [{"role": "user", "content": "Return ok."}]
  }'

Standard SDKs such as OpenAI Python/JavaScript reach the edge auth gate. Raw HTTP stacks that trigger Cloudflare 1010 should set a descriptive User-Agent.

Endpoint Auth Purpose Notes
GET /v1/models Generated key Authenticated model list for OpenAI-compatible clients. Use this to verify keys before wiring agents. Anonymous requests return guided 401.
POST /v1/chat/completions Generated key OpenAI-compatible chat completions with routing profiles and provider fallback. Supports common chat client flows, including streaming-shaped compatibility responses. tekizai/fusion is a premium chat model alias, and {"type":"tekizai:fusion"} is accepted as a premium server tool for Pro/Max-style synthesis.
POST /v1/responses Generated key OpenAI Responses-compatible shape for Codex and other tools that use the Responses API. Codex profiles should set wire_api = "responses". Pro and Max requests may attach {"type":"tekizai:fusion"} as a server tool.
GET /model-catalog Public metadata gateway-style discovery metadata for route families and model catalog pages. Discovery only. Live /v1/models and model traffic still require an active generated key.
GET /pricing Public metadata Hosted plans, limits, API base URL, checkout path, and public launch boundary. Managed model access remains disabled until provider permission, usage limits, and private audit records are ready.

Responses API

Use Responses wiring for Codex-style clients.

curl https://api.tekiz.ai/v1/responses \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tekizai/frontier",
    "input": "Return one sentence about TEKIZ.AI."
  }'
Signal Header or field When to use it
Authentication WWW-Authenticate, apiKeyPrefix, accountUrl, accountSetupUrl, primaryActionUrl, managedAccessUrl, keyRecoveryUrl Diagnose missing, inactive, malformed, or revoked generated keys, route no-key signups to generated-key setup, keep review-only one-subscription intake available, and keep same-email setup-key recovery available.
Billing or quota X-Quota-Limit, X-Quota-Used, X-Quota-Remaining, X-Quota-Period, X-Quota-Reset, upgradeUrl, billingUrl, supportUrl Show plan usage, monthly reset timing, and safe recovery links without exposing prompts or generated keys.
Rate limit Retry-After, X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset Back off safely when clients exceed request-per-minute limits.
Reliability X-TEKIZAI-Router-Retry-Count, status page health, selected upstream Understand failover behavior when a backend returns 401, 429, 502, 503, or 504.

Boundary

The hosted API is authenticated and customer-authorized.

  • Anonymous /v1/* model traffic is blocked with guided onboarding errors.
  • TEKIZ.AI does not grant unauthorized provider access or pool personal provider accounts by default.
  • Do not send prompts, provider credentials, OAuth tokens, generated keys, private keys, session cookies, raw provider responses, or customer data in support requests.