Appearance
Migrating from OpenRouter
Both are OpenAI-compatible gateways with <author>/<model> IDs, so most code moves by changing the base URL and key:
diff
- baseURL: "https://openrouter.ai/api/v1",
- apiKey: process.env.OPENROUTER_API_KEY,
+ baseURL: "https://api.usefully.info/v1",
+ apiKey: process.env.USEFULLY_API_KEY,Model IDs
Check each model you use against GET /v1/models. Common models use the same IDs (anthropic/claude-sonnet-5-5, anthropic/claude-haiku-5-5). Variant suffixes such as :free or :nitro aren't supported; use provider.sort instead (below).
Provider routing
Usefully's provider object covers the same ground:
| OpenRouter | Usefully |
|---|---|
provider.order | provider.order |
provider.allow_fallbacks | provider.allow_fallbacks |
provider.only / provider.ignore | provider.only / provider.ignore |
provider.sort (price, throughput, latency) | provider.sort (same values) |
provider.max_price (prompt, completion) | provider.max_price (input, output), USD per million tokens |
provider.data_collection: "deny" | provider.data_policy: "no_training" (or "zdr") |
models: [...] fallback list | Not supported yet; fallback is across providers of one model |
Provider IDs are lowercase (groq, together, anthropic); see each model's providers in the catalog.
Headers and extras
HTTP-RefererandX-Titlearen't needed and are ignored.- Beyond chat, the same key also does embeddings and web search.
Billing
Both are prepaid with pass-through token prices. Usefully adds a 3.5% fee on credit purchases and no per-request markup on tokens. See Billing.