gremlord wraps Claude Code in a thin local router. Same TUI, same tools, but the model behind it can be Anthropic, OpenAI, xAI, Ollama, or any OpenAI-compatible endpoint. Every token is priced, and hard budgets stop the spend. Free, MIT licensed.

Hey Product Hunt, Maor here - I built gremlord (a gremlin that minds your AI spend). Claude Code is the best agent harness I've used, but it only talks to Anthropic and it never tells you what a session cost. gremlord fixes both without forking it: a small local router points the unmodified claude binary at any provider (OpenAI, xAI, Ollama, vLLM, OpenRouter, anything OpenAI-compatible), prices every token as it streams, and enforces budgets that actually stop the next request instead of warning you after. My favorite part is the routing: a cheap classifier sends each turn to the smallest model that can actually hold it, so planning lands on a big model and mechanical edits land on a cheap one. Small numbers, honestly shared: no telemetry, 66 raw downloads mostly from my own updates, so ~10-20 genuine installs I can verify - but 56 unique git cloners in the last two weeks against 13 stars. People are trying it. It's free and MIT licensed: https://github.com/gremlord/grem.... Ask me anything - I'll be here all day.

That animation is freaking great. How did you make it?

As for gremlord: I'm interested to know how you managed to measure the differences of different models on the same harness. The claim that the harness is great but you wanted to try the different models, makes me believe you have a clear distinction on where the harness and the model begins.

I know in theory how this works but in practice because of very little visibility I don't *really" know what goes into the prompt.

That animation is freaking great. How did you make it?

I know in theory how this works but in practice because of very little visibility I don't *really" know what goes into the prompt.

gremlord

Claude Code on any model, with a budget

Claude Code on any model, with a budget