I built Faber because I kept seeing coding agents read far more of a repository than they actually needed for focused tasks, which slows things down and drives up token costs. Faber is free and open source; model API charges, if any, are billed directly by the provider you choose. The core idea was to use a code graph to identify the files, symbols, callers, and dependencies that matter first, so the model gets less unnecessary context and can respond faster. I then added prompt caching, token-conscious output, model switching between Claude, OpenAI Codex and local models, Git-aware workflows, memory, and a /usage dashboard that shows exactly where tokens, cache savings, and cost are going. My goal with Faber is simple: make AI-assisted coding much faster and significantly cheaper by reading less, reusing more context, and only sending the model what actually matters.

Forum Threads

p/faber

•

1d ago

Can an AI coding agent be faster, cheaper, and more accurate by reading less code?

I built Faber around this question.

Most coding tasks only depend on a small part of a repository, but an AI coding agent can end up reading a lot of unrelated code. That means more tokens, higher API cost, more latency, and potentially more irrelevant context for the model.

View all

Faber

AI coding agent with code-graph search and lower token cost

AI coding agent with code-graph search and lower token cost

Forum Threads

Can an AI coding agent be faster, cheaper, and more accurate by reading less code?

Forum Threads

Can an AI coding agent be faster, cheaper, and more accurate by reading less code?