Linkrra Log in Sign up free

Guide · Switch from OpenRouter

Switch from OpenRouter to Synap

4 steps · about 2 minutes

  1. Get a free key

    Sign in with Google at synap.linkrra.com — no card. Your key starts with sk-pool-.

  2. Swap the base URL and the key

    Both APIs are OpenAI-shaped, so these two values are most of the migration.

    export OPENAI_BASE_URL=https://synap.linkrra.com/v1
export OPENAI_API_KEY=sk-pool-...

  3. Keep or shorten the model id

    Most ids are byte-identical. A few lose their vendor prefix. Read /v1/models for the live list rather than copying a table.

    openai/gpt-oss-20b                 ->  gpt-oss-20b
google/gemma-4-26b-a4b-it          ->  gemma-4-26b-a4b-it
qwen/qwen3-coder-30b-a3b-instruct  ->  qwen3-coder-30b
qwen/qwen3-vl-30b-a3b-instruct     ->  unchanged
z-ai/glm-5.3                       ->  unchanged

  4. Drop the routing fields

    transforms, provider, models and route are accepted and ignored — no error is raised. Use the synap-v1 id if you want automatic routing.

    // before
{ "model": "...", "provider": { "order": [...] }, "route": "fallback" }

// after
{ "model": "synap-v1" }
Notes, prices and limits

Both APIs are OpenAI-shaped; the base URL and a few ids change. Recommended model: qwen3-coder-30b — $0.0675 in and $0.27 out per 1M tokens, 98,304-token context.

Checked live against both catalogues on the day of writing: /v1/models and OpenRouter’s, id by id. Every self-hosted id in the mapping above answered 200.

The OpenRouter-only fields are the trap, and they are silent: a request carrying transforms, provider.order, a models fallback array and route returned 200 and simply ran the model named in model. No error, and no fallback behaviour. Delete those fields rather than leave them in.

The HTTP-Referer and X-Title attribution headers are likewise accepted and ignored.

Routing happens by model id instead: synap-v1 picks the cheapest of the models we serve that can do the job, and bills at the price of whichever ran.

On the free tier: every request counts input and output against 100,000 free tokens a day, and answers stop at 4,096 output tokens. Coding agents send long prompts, so the free tier covers a short session; credits cover the rest.

Field names from the OpenRouter API reference, with both catalogues read live, checked 2026-11-03 20:20 UTC. Prices refresh from /v1/models when this page loads.