About

Evidence over anecdote.

Agent Tokenomics is an independent, open benchmark project. It exists to answer one narrow question well: how much does a development framework’s design actually cost an AI coding agent in tokens, using a methodology that anyone can inspect, dispute, and reproduce.

What this site is not

–Not a ranking of frameworks in general, only of measured token cost on specific, published task suites.
–Not a claim that lower token usage means a better framework (token cost is one input among many to that decision).
–Not final. Every number here is versioned to a model, a framework release, and a date, and is expected to be superseded as all three change.

Contact

Questions, corrections, or submissions: hello@agent-tokenomics.dev