Manifest

Cutting your token bill without rewriting your prompts is the job Manifest does. Manifest is an open-source LLM router built for OpenClaw that intercepts each query and sends it to the most cost-effective model for the complexity of the task.

The project says this can save up to 70% on token costs.

Scoring happens on your own machine and takes under 2ms, so routing adds little latency.

Privacy is part of the design: the cloud version uses a blind proxy architecture.

Manifest also lets you track costs, set usage limits and export data through the OpenTelemetry standard, which helps if you already collect metrics elsewhere.

It is geared toward OpenClaw users, so check that fit before you plan around it.