Per-turn speedometer for pi: TTFT, prefill tok/s, decode tok/s, and total wall time.

Live tok/s meter for OpenCode TUI with persistent avg/max/min summary after each response

Status bar tok/s + elapsed time for the Pi coding agent

OpenCode TUI plugin that displays real-time LLM output token speed (tok/s)

Hardware-aware CLI that scores your machine and tells you which local LLMs will run (yes/slow/no + tok/s) before recommending, installing, serving, and migrating them.

Can my machine run this model? Estimate VRAM/RAM, tok/s, and TTFT.

Plugin to add Multi-Collateral Dai support to dai.js

Pi extension showing subscription usage for supported model providers.

Self-learning LLM runtime — TurboQuant KV-cache (6-8x compression), SONA adaptive learning, FlashAttention, speculative decoding, GGUF inference

A pure JavaScript implementation of a BPE tokenizer (Encoder/Decoder) for GPT-2 / GPT-3 / GPT-4 and other OpenAI models

Self-Loading Universal Context Layer for AI Coding Agents. Stop teaching every AI your codebase. TokenCap makes project intelligence self-discoverable.

Enhanced Pi footer with git status, token usage, context usage, and model telemetry.

A CLI tool built by TOKTOKHAN.DEV

Fast WebGPU runtime for 1-bit (binary-weight) LLMs in the browser. Bit-exact, zero runtime dependencies.

💀 Find the agent skills you never use. Local-first, zero-dependency analytics for Claude Code & Codex skills: invocation counts, token cost, dead-weight detection.

Score how AI-readable a codebase is — offline, no keys needed.

Token and cost usage analytics command for Pi session history.

TypeScript Parser Combinator

Show LLM costs and token usage per agent in OpenCode sidebar

A CLI plugin for managing commits built by TOKTOKHAN.DEV