API troubleshooting

Fix hosted API errors without sharing secrets.

Find your HTTP status below. Check your key, plan, or service status, then use a short test request to confirm the fix.

Find your error Open account Check status

We will send a link to the email address you enter.

Hosted model traffic uses https://api.tekiz.ai/v1, generated sk_tekiz_* keys, and route profiles such as tekizai/frontier.

Endpoint

Check base URL

Use https://api.tekiz.ai/v1 for hosted OpenAI-compatible clients. Browser account UI lives on https://app.tekiz.ai, not the API host.

Key

Check prefix

Hosted model traffic requires an active generated sk_tekiz_* key. Full keys are shown once and should never be pasted into support.

Route

Check model profile

Start with tekizai/auto. Use your key with /v1/models to see which other profiles are available to your account.

Status What it means Useful response signals Next action
401 The key is missing, revoked, inactive, malformed, or not a TEKIZ.AI generated key. WWW-Authenticate, JSON accountUrl, accountSetupUrl, primaryActionUrl, managedAccessUrl, keyRecoveryUrl, pricingUrl, statusUrl, openaiBaseUrl, apiKeyPrefix, placeholder setupSnippet, and placeholder codexSetupSnippet. Confirm the client sends a generated sk_tekiz_* key. Create a new key if needed, then test /v1/models. Replace the placeholder in a copied setup snippet. Managed access needs review.
402 The account cannot route because payment is incomplete, the plan is inactive, or the quota is used up. X-Quota-Limit, X-Quota-Used, X-Quota-Remaining, X-Quota-Period, and account or pricing links in the JSON body. Open Account to check free requests, plan status, and current usage. Finish checkout or wait for the quota period to reset if needed. Contact support if a manual payment was confirmed.
429 The request-per-minute limit or abuse control was reached. Retry-After, X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset, and quota headers. Reduce concurrency, retry after the indicated delay, batch less aggressively, or upgrade to a plan with a higher rate limit.
403 / 1010 Cloudflare blocked the request before it reached TEKIZ.AI. JSON body with error_code: 1010 or browser_signature_banned; no X-TEKIZAI-Router-* headers. Use the OpenAI SDK, curl, or set a normal User-Agent for raw urllib-style clients; then retry /v1/models.
503 No approved model is available to serve this request right now. Status page health, selected upstream, backend health checks, and any safe request id or route profile name. Check public status and retry after service returns. If you run a private router, check that its models are healthy.

Safe probes

Check the response headers, then send one short request.

If raw Python urllib gets Cloudflare 403 / 1010, use a normal client User-Agent so the request reaches the TEKIZ.AI auth gate. 

curl -i https://api.tekiz.ai/v1/models \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "User-Agent: OpenAI-compatible-client/1.0"

curl -i https://api.tekiz.ai/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "User-Agent: OpenAI-compatible-client/1.0" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tekizai/auto",
    "messages": [{"role": "user", "content": "Return ok."}]
  }'

Python urllib

Check authentication from a plain Python client.

The placeholder key should return 401 from TEKIZ.AI. Replace it with a generated sk_tekiz_* key and retry. A Cloudflare 1010 response means the request did not reach the TEKIZ.AI API. 

import urllib.error
import urllib.request

req = urllib.request.Request(
    "https://api.tekiz.ai/v1/models",
    headers={
        "Authorization": "Bearer sk_tekiz_REPLACE_WITH_GENERATED_KEY",
        "User-Agent": "OpenAI-compatible-client/1.0",
        "Accept": "application/json",
    },
)

try:
    with urllib.request.urlopen(req, timeout=15) as response:
        print(response.status, response.read().decode("utf-8", errors="replace"))
except urllib.error.HTTPError as exc:
    print(exc.code, exc.read().decode("utf-8", errors="replace"))

401 setup snippet

Start with a placeholder, then add your key.

Replace only the placeholder key after the account page creates your generated sk_tekiz_* key. Keep real bearer tokens out of support threads and screenshots. 

# TEKIZ.AI hosted setup
export OPENAI_BASE_URL=https://api.tekiz.ai/v1
export OPENAI_API_KEY=sk_tekiz_REPLACE_WITH_GENERATED_KEY

curl "$OPENAI_BASE_URL/models" \
  -H "Authorization: Bearer $OPENAI_API_KEY"

Codex 401 profile

Use the hosted frontier profile after key setup.

A 401 response also includes codexProfile and a placeholder codexSetupSnippet. Use them to restore Codex or OpenClaw without guessing the provider, profile, model, or API type.

Codex setup guide 
# TEKIZ.AI hosted Codex setup
export TEKIZAI_API_KEY=sk_tekiz_REPLACE_WITH_GENERATED_KEY
mkdir -p ~/.codex
# Add this provider block once to ~/.codex/config.toml.
cat >> ~/.codex/config.toml <<'TOML'
[model_providers.tekizai]
name = "TEKIZ.AI Hosted"
base_url = "https://api.tekiz.ai/v1/"
env_key = "TEKIZAI_API_KEY"
wire_api = "responses"
TOML

# Create ~/.codex/tekizai-frontier.config.toml.
cat > ~/.codex/tekizai-frontier.config.toml <<'TOML'
model_provider = "tekizai"
model = "tekizai/frontier"
TOML

# Plain codex keeps using ChatGPT. TEKIZ.AI is explicit.
codex --profile tekizai-frontier

What to send support

Share error details, not private data.

Send the HTTP status, time with timezone, request ID if present, route profile, and the result of the Account key test. Leave out prompts, credentials, full keys, and customer data.

Technical response details that can help
  • Keep non-secret headers such as WWW-Authenticate, Retry-After, X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset, X-Quota-Limit, X-Quota-Used, X-Quota-Remaining, and X-Quota-Period.
  • Keep JSON onboarding links such as accountUrl, accountSetupUrl, primaryActionUrl, managedAccessUrl, keyRecoveryUrl, pricingUrl, statusUrl, openaiBaseUrl, the non-secret apiKeyPrefix, and placeholder setupSnippet / codexSetupSnippet. Remove full bearer tokens and prompt bodies.

Contact support

Provider access

Use only provider access you are allowed to use.

Hosted plans include inference on available hosted capacity, account management, API keys, usage limits, and request history. Use models available to your account. To use your own provider credentials, run a router you control. Managed model access is available after provider and plan approval.

Request one-subscription beta