Frequently asked questions

One OpenAI-compatible API in front of every major LLM provider. You get a single key and a single prepaid bill across all of them, plus team roles, spend controls, and a complete audit trail. Point your existing code at it and call any model by a provider/model id.

No. Point your existing OpenAI SDK at your workspace's /v1 base URL and use an wor_live_… key. The request shape, response shape, streaming, and tool-call schema are all identical — every language and library that already speaks OpenAI just works. The How To Configure page has copy-paste snippets.

OpenAI, Anthropic, Google Gemini, Groq, self-hosted open models via Ollama, and any OpenAI-compatible endpoint. Call a model by its id — e.g. openai/gpt-4o, anthropic/claude-sonnet-4.6, or gemini/gemini-2.5-pro. The live catalog and its per-million-token prices are on the Models page in your workspace, and new models appear the moment they're added — no SDK update needed.

Change the model string. That's it — no SDK swap, no provider-specific retry logic, no quirks. Streaming and OpenAI-shaped function calling work the same way across every provider; we translate to each one's native format.

Anything that lets you set a custom OpenAI-compatible base URL and an API key. Known to work out of the box: LiteLLM, Open WebUI, LobeChat, BoltAI, llmcord, vLLM/Helicone proxies, and anything built on the openai or openai-node SDKs.

Sign up for a workspace, confirm your email to claim $1 of starter credit, mint an API key on the API Keys page, and point your client at your /v1 base URL. You're live in a couple of minutes. Create a workspace.

Pass-through usage pricing: you pay each provider's own per-token rate — there's no markup on usage. Every model's price is shown per-million-tokens in the live catalog, so there are no hidden conversion rates.

A flat 5% platform fee on credit top-ups — a $100 deposit adds $95 of spendable credit. That's the only fee; the usage itself is billed at cost.

Prepaid credit via Stripe Checkout. You top up once, and every call across every provider debits the same balance — no reconciling four separate provider invoices.

Calls return a clean 402 instead of silently overspending. Your team knows immediately and finance never gets a surprise bill. Free self-hosted models remain callable even at a zero balance.

Yes — optional automatic top-up charges your saved card whenever the balance falls below a level you choose (the same 5% fee applies). It turns itself off after 3 consecutive failed attempts so a dead card can't loop.

Every new workspace starts with $1 of starter credit, and free self-hosted models can be called even at a zero balance. You only pay when you top up to use the paid providers.

In real time. The dashboard shows daily spend, request count, and remaining credit, and the activity log records which model each request hit and what it cost — no waiting until end of month.

Yes. Invite teammates with a shareable one-click link and a pre-set role. Links expire in 7 days and are revocable anytime.

Four roles — Owner, Admin, Member, Viewer — plus custom permission groups (e.g. "Finance reviewers" with billing access only) and Departments for org structure. Admins manage members and keys; viewers see usage but can't spend.

Keys are stored as one-way SHA-256 hashes — the plaintext is shown once at creation and never again. Each key can carry its own requests-per-minute limit and be assigned to a department, so a runaway script can't drain the whole workspace overnight.

Yes. Connect your workspace to your own identity provider — Microsoft Entra, Okta, Keycloak, or any OIDC provider — and your team signs in with company credentials. It's configured self-service under Settings → SSO.

Yes — native SCIM 2.0 endpoints per workspace for SailPoint, Okta, and Entra. Your IGA system can provision and de-provision users, sync group memberships, and run access certifications; when a user is deactivated upstream, their workspace access disappears within seconds.

Yes — worouter is also an MCP server. Add your workspace's /mcp URL as a custom connector in Claude Desktop; it authenticates over OAuth and exposes list_models and chat tools, billed to your workspace like any other call.

From your workspace to whichever LLM provider you call — nothing else. Every workspace is fully isolated: its own subdomain, credits, keys, and audit trail. Workspaces never see each other.

Yes — every credit movement and every API call is recorded with attribution (model, cost, latency, and which key made it). It's useful for SOC 2 or ISO 27001 evidence, or just answering "who spent that $40?"

Yes. worouter is a single Go binary plus Postgres — run it in your own Kubernetes for data residency, or use our hosted service. Either way, data goes only from your network to the provider you choose.

The gateway runs multi-replica with health checks and zero-downtime rolling releases, and every production build is a tagged git commit you can trace. Structured JSON logs on every request (workspace, user, key, model, cost, latency) pipe straight into Splunk, Datadog, or whatever your SIEM is.