Migrate from OpenRouter

OpenRouter and Synapse Garden share the same creator/model-slug model id format and OpenAI-compatible wire shape. Migration is just a base URL and an API key — and the same swap works for the AI SDK's streamText when you want a more idiomatic TypeScript surface.

Diff

  import OpenAI from "openai"

  const client = new OpenAI({
-   apiKey: process.env.OPENROUTER_API_KEY,
-   baseURL: "https://openrouter.ai/api/v1",
+   apiKey: process.env.MG_KEY,
+   baseURL: "https://synapse.garden/api/v1",
  })

  await client.chat.completions.create({
    model: "anthropic/claude-opus-4.6", // ← same format, no change needed
    messages: [{ role: "user", content: "..." }],
  })

Model names with creator/slug syntax — already what you're using on OpenRouter — pass straight through.

What changes

What you gain

• Multiple compat surfaces. OpenAI and Anthropic SDKs both work — pick whichever your team prefers.
• Per-project governance. OpenRouter is account-scoped; we're project-scoped. Spend caps, allowlists, rate limits are all configurable per project.
• Audit log. Every key action and mutation is logged. Useful for SOC 2, internal compliance, postmortems.
• Predictable plan + overage — flat included tokens per plan, overage at list rate. No "credits expire" surprises.

What stays the same

• Streaming.
• Tool use.
• Provider routing semantics — providerOptions.gateway.order, only, sort all work the same way.
• Model fallback chains — providerOptions.gateway.models works the same.
• Vision, JSON mode, function calling — all unchanged.

Migration steps

- OPENROUTER_API_KEY=sk-or-v1-...
+ MG_KEY=mg_live_...

Provider routing translation

Both platforms use providerOptions.gateway.{order, only, sort, models} — drop your OpenRouter routing config in unchanged:

providerOptions: {
  gateway: {
    order: ["bedrock", "anthropic"],
    sort: "cost",
    models: ["openai/gpt-5.4-mini"], // fallback chain
  },
}

Resolution order is identical: order first, sort for the rest, only filters, models provides cross-model fallback.

Pricing

Compare your OpenRouter bill (per-token rate × usage) against our /pricing page. For most accounts, Synapse Garden is comparable on the per-token rate, with a small included-tokens floor at the plan tier.

If you're a heavy OpenRouter user (>10M tokens/mo), our Pro ($50/mo, 100M included) or Agency ($200/mo, 500M included) plans usually come out ahead.

Caveats

• Provider preferences — OpenRouter has its own provider preference UI; ours is the same providerOptions.gateway.* API. Per-account preference UI is on the dashboard roadmap.
• Model availability — every model in OpenRouter's catalog is in ours, but new models can land at slightly different times. Check /models for current availability.
• Free tier — both platforms have free tiers; ours is plan-based (1M tokens/mo) rather than per-model.
• Anthropic SDK compat — OpenRouter doesn't speak the Anthropic Messages format directly; we do (https://synapse.garden/api with the Anthropic SDK).