SharedLLM routes Claude Code and friends through whatever model you choose — with smart auto-routing, live usage metering, and per-developer spend control. Bring your own keys, or tap the shared pool.
curl -fsSL https://sharedllm.com/install.sh | shmacOS / Linux — auto-detects bun · pnpm · yarn · npm, or bootstraps bun.
Ships both sllm and sharedllm. Then sllm login → sllm claude.
One command across your whole stack
Keep the agents and keys you already use. SharedLLM sits in the middle and makes everything routable, observable, and accountable.
Launch Claude Code today — Codex and Aider next — through a single command. The CLI is a launcher, not a proxy: it sets the right env vars and execs your tool.
Auto resolves each request to the first model with an available key — your own contributed key first, then the shared pool. Set your priority list, up to 10 deep.
Anthropic, OpenAI, Ollama, Kimi and more — native or via transparent Anthropic↔OpenAI translation. Swap with a flag, no config rewrites.
Use the provider keys you already have. Encrypted at rest, scoped to you, and never shared without your say-so.
Every request metered. Track spend across providers, set per-developer caps, and watch routing happen live on the dashboard.
Share spare capacity with your team or the community — and draw from the pool when your own keys run dry.
Three commands from zero to a fully metered agent.
One line to install, one command to authenticate. A CLI key is created and stored with 0600 perms.
npm i -g sharedllm sllm login
Choose from the interactive picker, name one directly, or let Auto route to the best available key.
sllm claude --model auto
Your agent runs through the gateway. Every token is tracked, capped, and visible on the dashboard.
sllm claude # just works
The dashboard streams routing decisions as they happen — which agent called, which key answered, native or translated, and what it cost. No more guessing where your spend went.
Free to start. Self-host the open-source core, or use the hosted gateway. Your keys, your models, your limits.