omp packages

Project-scoped file memory for Pi coding agent sessions

Drop Anthropic thinking blocks that a system prompt or tool change has already invalidated

Pi runtime telemetry: compaction, tools, vault, skills, follow-ups, subagents -> NDJSON shards, HTML dashboard, agent tool

Scan Agent Skills for prompt injection, hidden instructions, exfiltration, and remote code before Claude Code, Codex, OpenCode, Pi, or npx skills installs them. Standalone and offline by default; TypeSafe's jev is an optional second opinion.

This extension is designed for one purpose: to capture a session’s continuation context more precisely than compaction and make it available to a fresh, natively linked Pi session through a local handoff file referenced by its first prompt. How work conti

Knowledge and Vault tools for native Pi hosts

User-managed output capture and recovery for Pi

Workspace-first authorization, scoped approvals, and macOS shell containment for Pi (formerly pi-warden). See SECURITY.md and docs/COMPATIBILITY.md for verified versions and limits.

Pi / omp package: first-party skills and themes

Pin a session goal and keep the agent anchored to it: settled-idle continuation, safety limits, evidence-gated completion

Pi extension package for Gemini OAuth compatibility

Emacs-style visual undo-tree for the Pi coding agent

Pi extension that discovers markdown rules and injects conditionally matched ones into the system prompt. Functional core, imperative shell.

Keep Pi running through LLM-relay 400s: retry errors its substring classifier misses, and strip replayed thinking signatures a pooled relay rejects.

Realtime voice and LAN remote control for any Pi model

Caps Pi's compaction file lists (<read-files>/<modified-files>): drops dead, duplicate and temp paths, hides sensitive ones, strips blocks the summarizer re-seeds into the narrative, and keeps the cost per turn constant.

Pi extension that bundles a model, thinking level, tools, and system prompt into switchable presets

Pi extension for Superpowers workflows: recon, research, implementation, review, and debug

Privacy-conscious OpenTelemetry traces, metrics, and optional LLM evaluation for Pi

Pi extension that checks installed plugin versions against npm and updates them safely on request

Semantic search over local files for pi. Indexes a directory of text files, watches for changes, and exposes a knowledge_search tool to the LLM.

A chartered project-manager agent plus the specialist team it delegates to: kernel/charter pm, pack-based intake interview, researcher, oracle, worker, and the ship-gate reviewers. Bundles the subagent runtime and the execution pipeline.

On-demand Roblox Studio MCP tools for Pi without a long-running MCP server.

Image generation and reference-image editing for pi.

woku AI Toolkit: connect your agent to woku's voice-of-customer platform. Design VoC programs and turn customer feedback into support tickets and action plans, through woku's MCP server.

Your constellation of enduring experts for Pi

Minimal plan mode for Pi — read-only exploration and planning before implementation, with dedicated model support.

Writes a small set of memory candidates when Pi exits, then leaves the final decision to you. There is no database, vector store, or hidden profile: candidates go to a plain Markdown inbox for manual review.

Inference speed extension for pi-agent: tokens/sec and time to first token.

Auto-continue on empty model responses + configurable timeouts for bash and search tools

A pi extension for stashing draft messages and restoring them later.

Pi extension — broker .env secrets to tools via $KEY references with gated injection

Pi agent extension: post to a GoToSocial instance as an evolver identity. Performs the 157 principal-login ceremony with the agent's OWN key, then uses the ordinary Mastodon-compatible API.

Session recap for Pi — shows a concise summary of what's happening above the TUI input box, like Claude Code's ✦ Recap: feature

Auto-switch pi between a primary and fallback model based on quota, with /usage and /auto-model commands.

Token-efficient autonomous task execution with context collapse

HexaSync AI toolkit: agent skills for HexaSync profile templates plus the HexaSync Template IDE MCP server, installable into Claude Code, Codex, Cursor, Antigravity, pi, OpenClaw and VS Code.

Autonomous goal loop for pi: an AI judge decides whether the goal is done, otherwise the original goal is re-sent to the agent.

Set per-model automatic compaction thresholds with /autocompact

RENAMED to agents-cron — keep this stub only so old `pi-scheduler` crontab entries and installs keep working.

Persistent cross-session prompt history for Pi.

Slack tools for Pi that act as YOU, not as a bot. Adds 8 LLM-callable tools (slack_*) that read channels, threads, DMs, and search results, plus post, update, and delete messages using a Slack USER token (xoxp-). No bot to invite, no footprint in the work

Respan instrumentation plugin and pi extension for the pi coding agent (@earendil-works/pi-coding-agent)

OpenCode provider for Pi Agent. Auto-discovers models from OpenCode registry and uses Pi's native OpenAI-compatible handler.

Sync your ~/.pi/agent config across machines through a private git repo — with hard guards that keep secrets out.

Unofficial Pi provider extension for Volcengine Ark Agent Plan.

Free frontier models (DeepSeek V4.1 Flash, Kimi K3, GLM-5.3, ...) for the pi coding agent via a self-contained local OpenAI-compatible gateway. Ported from Ebony-Vinyl/dsh-our-free-model.

Zero-model-call Pi commands for Agent Golf login, raw session submission and problem browsing

Baseline Pi behavior for session naming, history, prompt editing, fullscreen navigation, and footer styling

Compact, theme-aware rendering for pi's built-in tools: one-line summaries, syntax-highlighted reads, colorized diffs — behavior untouched