Small, sharp tools for building with LLMs.

size-limit for LLM prompts. Count the tokens in your prompt and context files, set budgets, and fail CI when one grows past its limit.

Make LLM output pass muster. A tiny verify-and-retry loop: generate → check (schema / predicate / LLM-judge) → retry with feedback until it passes.

Instant, offline token + cost counter for LLM text. cat a file or pipe stdin and get the count — plus a cost estimate across models.

Talk to an LLM from your terminal. Pipe context in — git diff | heyllm "write a commit message" — over any OpenAI-compatible API.

Validate your .env against an .env.example contract — flags missing, extra, and empty keys. Runs locally and as a CI gate.

Interactively sweep stale local git branches — the ones already merged or whose remote is gone. Daily git hygiene, solved.

Watch your AI agent think. A 1-line wrapper records its run; replay it as a live terminal dashboard of tool-calls, tokens, and cost.

Turn a Markdown file into a gorgeous animated terminal demo (SVG) — from the shell snippets you already wrote. No tape files.

Generate a stunning, shareable SVG card for any repository — languages, stars, description — for social previews and READMEs.