Give your host a workforce.
Leverage speaks MCP in both directions: your host calls it to run missions, and it calls back to your host to borrow the model your subscription already pays for.
Install
claude mcp add leverage -- node /abs/path/to/leverage/mcp/server.ts
{
"mcpServers": {
"leverage": {
"command": "node",
"args": ["/abs/path/to/leverage/mcp/server.ts"],
"env": { "LEVERAGE_API_URL": "http://localhost:3000" }
}
}
}The server talks to a running Leverage instance, so start it with npm run dev first. Then, inside your host:
Use Leverage. Finish this application. Budget $0. Quality production.
The five tools
Five, not forty. The host is a strategist, and a strategist needs to state an outcome, watch it, stop it, and inspect the evidence. Everything else is Leverage’s job.
Lending your seat back
If your host offers the sampling capability, the Leverage MCP server registers itself as a worker source. Leverage can then hire your host’s own model through sampling/createMessage. No API key is minted, none is stored, and the inference is billed to the seat you already pay for.
The control plane cannot initiate sampling. Only the process holding the protocol connection can, so the plane parks requests on a queue and the MCP server drains them: register, claim, sample, post back.
leverage: Host seat registered. Leverage can now hire your own model with no API key.
Budget behaviour
leverage_run defaults to budgetMaxUsd: 0, which is a hard block rather than a preference. Paid models never enter the ranking pool. Host and local workers are still eligible, because a subscription seat and a local runtime are not Leverage spending money.
{
"goal": "Finish the application and verify it",
"budgetMaxUsd": 0,
"qualityTarget": 0.95,
"privacy": "prefer-local",
"parallelism": 2
}