Use any model. Pay one bill. Control who uses what.

Point your existing OpenAI SDK at your workspace and you're done. Same request shape, same response shape, same streaming, same tool-call schema — every language and library that already speaks OpenAI just works.

Change one string from openai/gpt-4o to anthropic/claude-sonnet-4.6 or gemini/gemini-2.5-pro — no SDK swap, no retry logic, no provider-specific quirks.

One contract across the major closed models and your own open-source deployments — including air-gapped Ollama if your data can't leave your environment.

Server-sent events and OpenAI-shaped function calling work the same way across every provider; we translate to each one's native format.

Browse every model and its per-million-token price from one page. New models show up the moment they're added — no SDK update needed.

You top up credit once via Stripe; we route to whichever LLM you call and debit against the same balance. No more reconciling four provider invoices.

Cap requests-per-minute on each API key. A misbehaving script can't drain the workspace credit overnight.

When your credit hits zero, calls return a clean 402 instead of silently overspending. Your team knows immediately; finance never gets a surprise.

Every workspace sees its daily spend, request count, and remaining credit on the dashboard. No waiting until end-of-month.

Buy credit on a card via Stripe Checkout. No contract negotiations to add $100 of usage for a hackathon.

Owner / Admin / Member / Viewer — pick what fits each teammate. Admins can manage members and keys; viewers see usage but can't spend.

Build groups like "Finance reviewers" (billing access only) or "Engineers" (mint keys + revoke). Mix and match to fit your real org structure.

Group people by Engineering, Marketing, Finance — keep your member list organized as your team grows.

Send a one-click join URL with a pre-set role. Expires in 7 days; revocable anytime.

Recent activity shows which model each request hit and what it cost. Find the team driving spend at a glance.

Native SCIM endpoints per workspace — your IGA system can provision and de-provision users, manage group memberships, and run access certifications without us building a custom connector.

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

When SailPoint deactivates a user, their workspace access disappears within seconds — no orphaned accounts, no quarterly cleanup spreadsheet.

Add or remove a member from a SailPoint group and their worouter permissions update automatically. Birthright + access reviews stay in your governance system.

Every customer is fully isolated — its own subdomain, its own credits, its own keys, its own audit trail. Workspaces never see each other.

Runs in your own Kubernetes or ours. Data goes from your network to whichever LLM provider you choose — nothing else.

API keys, SCIM tokens, and passwords are stored as one-way hashes. The plaintext is shown once at creation and never again.

Every credit movement and every API call is recorded with attribution — useful for SOC 2, ISO 27001, or just answering "who spent that $40?"

The whole interface is server-rendered — no opaque JavaScript bundles handling your tokens in the browser. Less attack surface, easier to audit.

Multi-replica with health checks, rolling updates, zero-downtime releases.

Pin a version, ship deliberately. Every prod image is a tagged git commit you can trace.

JSON logs on every request with the workspace, user, key, model, cost, and latency. Pipe straight into Splunk, Datadog, or whatever you already use.

Add a new model or change a price from the admin UI — running services pick it up within a minute, no rolling restarts.