LLM Gateway
One gateway for every model provider you run
memtrunk sits between your tools and your LLM providers — issuing keys, pooling accounts, routing requests, and giving you a real accounting of what it costs, who used it, and why.
What's inside
Provider & Key Management
Register every upstream provider once, hold its credentials centrally, and issue scoped API keys to the clients that call through you.
Account Pooling
Draw from a pool of upstream accounts so load and quota spread automatically instead of hitting one account's ceiling.
Intelligent Routing
Route every request to the right provider and model — including rules that span providers — without touching client code.
Usage & Cost Telemetry
See tokens, cost, and volume as they happen, and enforce spend and rate ceilings per key, org, or team before they become a surprise bill.
More capabilities
Why memtrunk
Subscription pooling, not just metered keys
Unlike typical general-purpose LLM gateways, which broker metered API keys, memtrunk also pools real provider subscriptions — the team plans your organization already pays for — and spreads load across the pool automatically.
Governance included, not paywalled
Most self-hosted gateways treat role-based access and audit logging as a paid enterprise upgrade, even when the core gateway itself is free. In memtrunk, role-based access and an immutable audit trail ship as part of the console — nothing to unlock separately.
Per-agent telemetry, not just aggregate cost
Most LLM gateways stop at aggregate cost and usage tracking. Seeing which specific agent or tool call failed, and why, is usually a separate observability platform bolted on afterward — in memtrunk, that detail is tracked natively, per agent and per tool call.