Get started

Two moves: see what you'd save, then turn it on.

You don't need any of our infrastructure to find out if Wendl is worth it. Start with a breakdown of what you're spending now — routing comes after the numbers make the case.

Step 1 · no setup required

See what you'd save first.

Send us a slice of your existing usage — an OpenTelemetry trace export or a provider usage log — and we'll show you, on your own traffic, what routing through Wendl would cost versus what you pay now. No proxy, no install, no commitment.

Get my breakdown → See a sample breakdown →

Step 2 · turn on routing

Routing is what cuts the bill.

Wendl puts a small router in front of your models so routine work runs free on local models and the expensive ones get called only when they earn it. Pick how you want to run that router:

Managed — we run the router

early access

You don't touch LiteLLM, Docker, or servers. We host the router and the local models, connect to your own model keys where you want them, and hand you an endpoint + key to paste into the plugin. We also keep it tuned as new models ship.

Best if you'd rather not run infrastructure — this is the paid, done-for-you path.

Join early access →

while we onboard the first teams by hand — email gets you in the queue.

Self-host — run the router yourself

early access

The Wendl gateway is a single Node process — no Docker, no Postgres. Free to run; your data and local models never leave your box. The repo is early-access while we work with our first teams — request an invite and you're two commands from running:

# 1. from your checkout — start the gateway (no Docker) make bootstrap-lite # seed team + ops budget keys — prints them make gateway # serves :4000

# 2. point your agent at it — drop-in on both API shapes:

# nanoclaw / Claude Agent SDK (Anthropic Messages) ANTHROPIC_BASE_URL=http://localhost:4000 ANTHROPIC_AUTH_TOKEN=<team key from bootstrap-lite> ANTHROPIC_MODEL=tier-mid # OpenClaw plugin → plugins.entries.wendl.config "litellmBaseUrl": "http://localhost:4000", "litellmMasterKey": "${LITELLM_MASTER_KEY}"

Cloud tiers need an ANTHROPIC_API_KEY (and/or OPENROUTER_API_KEY). Local $0 tiers need native Ollama on :11434 — or just route cloud-only. Prefer the full LiteLLM + Postgres stack? make bootstrap still runs it with Docker.

best if you're technical and want to own the stack. Full config in the repo README.

Already running LiteLLM

available now

Point the plugin straight at your existing proxy — nothing new to stand up. You just need the base URL and a master key.

// in ~/.openclaw/openclaw.json → plugins.entries.wendl.config { "litellmBaseUrl": "http://localhost:4000", "litellmMasterKey": "${LITELLM_MASTER_KEY}" }

best if you already run LiteLLM. On to Step 3.

Step 3 · read it in chat

Put the numbers in your chat.

Install the free OpenClaw plugin and it prints your caps, spend, and savings right where you work — no agent call, no tokens burned.

$ openclaw plugins install git:github.com/wendl-ai/openclaw-wendl@main $ openclaw plugins enable wendl

Then type /limits, /stats, or /wendl. Full plugin docs are in the plugin repo.