Deterministic cl100k_base token profiler, priority ranker, and context compiler pipeline for PAKB

OpenAI cl100k_base vocabulary for hypertok format version 1

BPE tokenizer data files for LLMs (o200k_base, cl100k_base, DeepSeek V3/V2)

offline-capable ESM module for cl100k_base tokenization in the browser

A pure JavaScript BPE tokenizer with model metadata and pricing for GPT-5, GPT-4o, o-series, and other OpenAI models

CodeMode for Pi and OMP: read, edit, write and bash, with transactional files, source views and shared-input program batches.

_(inspired by [tiktoken-node](https://github.com/ceifa/tiktoken-node))_

Tokenometer CLI — LLM token cost + latency benchmarking across Claude, GPT-4o, Gemini, Mistral, and Cohere. Multi-format, empirical mode, vision tokens, SARIF output.

Prompt compression with tokenizer-aware savings checks and conservative literal protection.

A local-first AI plugin for Joplin that clusters notes semantically, suggests tags and notebook structures.

tiktoken Tokenizer adapter for token-budget, backed by pure-JS js-tiktoken by default (with an optional Node-only native/WASM path).

serena MCP (38% less tokens). Quick setup: npx serena-slim --setup

30 tools vs 1: an agent-payment MCP that measures 230 tokens of schema floor instead of 3,705 (16x smaller; ~48x effective across a multi-hop payment flow). Signed, sanctions-screened, trust-gated payments across x402, Stripe, AP2, MPP, Mastercard Agent P

Token budget linter for agent skill files

A pure JavaScript implementation of a BPE tokenizer (Encoder/Decoder) for GPT-2 / GPT-3 / GPT-4 / Claude Instant / Claude 2

Lazy-loaded common tokenizer presets for FFBPE

Core parser, types, and utilities for Agent Skills (SKILL.md)

A meticulously deterministic, exhaustively composable, intentionally maximalist tooling artifact for the comprehensive amplification of token expenditure across natural-language workloads — accompanied by a rigorously empirical, corpus-grounded, tokenizer

Turn noisy CLI output into compact structured results (fewer tokens), full logs preserved.

Tokenectomy Sentinel — Cut 95% of useless framework noise and token bloat from error logs in Claude & Cursor