Route every LLM request through a single gateway with authentication, cost tracking, rate limiting, and policy enforcement. Support 13+ providers out of the box.
Claude, GPT, Gemini, Mistral, Llama, Bedrock, DeepSeek, Grok, Perplexity, and more — all through one endpoint.
User keys (gk_user_*), service keys (gk_svc_*), and temporary keys (gk_tmp_*) with scoped permissions.
Real-time metering of every request. Track costs per agent, per task, per model. Set budgets and alerts.
9 PII detection patterns, secret redaction, content policy enforcement on both requests and responses.
An agent or user sends an LLM request via the gateway.
Gateway validates the key, checks policies, and enforces limits.
Request is routed to the configured provider with failover.
Response is metered for cost, logged for audit, and returned.