Gateway migration

Move OpenAI-compatible clients to TEKIZ.AI.

Create a generated sk_tekiz_* key, activate your free evaluation or a plan, and check that a model route is available. Then change the base URL in one client and test tekizai/auto. Use /v1/models to see which other profiles your account can run.

Create hosted API key Quickstart API reference Compare gateways Check status

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

Connect provider access you are authorized to use. Personal accounts stay with their owners. Managed provider access needs separate approval before activation.

Existing gateway setupTEKIZ.AI hosted setupMigration note
OPENAI_BASE_URL=https://gateway.example/api/v1 OPENAI_BASE_URL=https://api.tekiz.ai/v1 Most OpenAI-compatible SDKs only need the base URL changed.
LEGACY_GATEWAY_API_KEY or a Bearer key OPENAI_API_KEY=sk_tekiz_your_key_here Generate the TEKIZ.AI key on the hosted account page. Raw keys are shown once.
Provider-prefixed model IDs from the hosted catalog Start with tekizai/auto; select Frontier or Fusion only when your authenticated model list shows capacity. Provider-specific routes need authorized access on your TEKIZ.AI profile. Fusion can run separate tasks in parallel and review the result before returning one response.
Gateway /models Public https://api.tekiz.ai/model-catalog and authenticated GET /v1/models The public catalog lists examples. Your key shows which models your account can request.

Before

Typical gateway environment.

export OPENAI_BASE_URL=https://gateway.example/api/v1
export OPENAI_API_KEY=$LEGACY_GATEWAY_API_KEY

After

Hosted TEKIZ.AI environment.

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

First request

Verify routing before moving production traffic.

Send one short request after your account has active routing and an available model. Move agents and job queues after you confirm the key, plan, request limit, and route.

curl https://api.tekiz.ai/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tekizai/auto",
    "messages": [
      {"role": "user", "content": "Reply with one migration check."}
    ]
  }'
OpenAI SDK

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
  baseURL: "https://api.tekiz.ai/v1",
});

await client.chat.completions.create({
  model: "tekizai/auto",
          messages: [{ role: "user", content: "Reply with one sentence." }],
});
Routes

Profile migration

  • Use tekizai/auto for the first hosted request, then consult authenticated /v1/models for available profiles.
  • Select tekizai/frontier or tekizai/fusion only when your account's model list shows the profile. Fusion is available on Pro and Max when its worker and review models are online. See the Fusion guide.
  • Use {"type":"tekizai:fusion"} for explicit Fusion tool calls. TEKIZ.AI handles that server tool before downstream provider routing and applies the same Fusion plan gate.
  • Use tekizai/local when the router has local Ollama routes available.
  • Use model catalog discovery before mapping provider-specific IDs.
Safety

Provider boundary

Connect only provider access you are allowed to use. Managed provider access is in private beta. We activate a provider route after authorization, margin checks, quotas, durable audit records, abuse controls, and acceptable-use enforcement are ready. See the provider terms.

StatusLikely migration causeNext action
401Missing, inactive, malformed, or revoked generated key.Generate or rotate an sk_tekiz_* key on the account page.
402Checkout, subscription, plan, or monthly quota blocks model traffic.Check your billing status and remaining quota. Complete checkout, wait for the quota reset, or choose a suitable plan.
429Request-per-minute limit or burst control triggered.Back off, reduce concurrency, or move to a higher plan.
503No healthy authorized model route is available.Check public status. If you use a private router, check its health, provider access, and local models.

For response headers, safe probes, and no-secret support context, use API troubleshooting.

Cutover checklist

Move one client at a time.

  1. Create a hosted account and generated sk_tekiz_* key.
  2. Activate the free evaluation or a plan. Check available hosted capacity, or connect provider access you are authorized to use.
  3. Set OPENAI_BASE_URL=https://api.tekiz.ai/v1 in one development client.
  4. Send a tekizai/auto chat completion, then consult /v1/models before choosing another profile.
  5. Confirm usage, quota, and status before moving agent jobs.
  6. Keep your existing gateway available until your selected TEKIZ.AI profile and its model route are healthy.