Local-first analytics for AI agent skills. Track usage, measure context budget, and prune what you don't use.

Slate: thread-weaving orchestration for the pi coding agent — bounded worker actions, compressed episodes, mandatory research/design/adversarial/track review workflow, configurable model failover, and orchestrator context-budget handoff.

Context-budget instrumentation core for LLM applications: normalized token accounting, effective-budget gauges, turns-to-exhaustion prediction, compaction and handoff decision policies, and a handoff fidelity protocol.

Keep Pi runs controlled with automatic context-budget wrap-up and kill thresholds.

p5.js-style canvas sketches for React/Next.js — headless hooks plus drop-in <CanvasSketch>/<ThreeSketch> components, SSR-safe and DPR-aware, with an optional Three.js engine behind its own subpath export and a shared WebGL context budget.

MCP server for agentfit: token-aware message truncation. Fit a chat history into the model's context budget with drop-oldest, drop-middle, or priority strategies.

TypeScript library and CLI for allocating and rebalancing an LLM context-window token budget across named zones (system, tools, retrieval, history, buffer).

Pi extension that adds a /context modal for context budget breakdowns.

A lightweight registry of LLM model information, like name and context sizes, for building AI-powered apps.

Claude Code plugin: compresses tool outputs, blocks wasteful prompts, manages context budget. ~42% token reduction.

Built-in HUD sections for pi-hud: clock, context budget, git status, and working directory. Registers via the hud_section EventBus contract.

OCI NodeJS client for Budget Service

Tokenizer-aware chat-history fitter. Trims, windows, or summarizes old messages so your conversation fits the model's context budget — with a full audit log of every kept, dropped, and summarized message.

AI DevOps Framework - AI-assisted development workflows, code quality, and deployment automation

Ambient HUD injected into the model's context on every LLM call. Live briefing notes (time, context budget, git, cwd) that refresh as assistant messages complete — kept off the transcript and out of the user's view.

text/x-diff (unified/combined diff) mimetype handler for plurnk-service. Hand-rolled line-budget scanner; no parser dependency.

Context budget management for AI applications. Visibility, governance, and health scoring for LLM context windows.

This is an internal utility, not intended for public usage.

View docs [here](https://radix-ui.com/primitives/docs/components/context-menu).

Context window management for LLM applications. Tool result truncation, conversation compression, and context budget estimation.