omp packages
-
Vertex AI Gemini grounded web search and research tools for Pi, using Pi's google-vertex provider
-
Bounded shell execution with persistent PTY sessions for Pi
-
Pi extension that prevents context-mode's synthetic routing message from superseding the latest user request.
-
Web search, page/PDF/GitHub/YouTube content extraction tools for pi (fork of pi-web-access).
-
Refuses a git commit on the trunk the repository declares, and checks every commit a push would publish for the declared commit convention and for a good signature, before the push runs. It keeps nothing in context on every turn. Distributed as a Pi packa
-
Make Pi easier to follow while it works: keep the conversation and Working... visible while quietly hiding tool chatter and optional thinking. Toggle it with /calm on, /calm thinking, and /calm off.
-
Robust PowerShell extension for pi on Windows — token-based bash-to-PS translation, background jobs, /ps clone starts pi, diagnostics
-
Delegate Pi tasks to the local Codex app-server harness.
-
Pi extension that shows live prompt-cache reuse and cost diagnostics above the editor.
-
Send a desktop notification through the session notification daemon, and notify when a response settles.
-
Pi skills, prompt templates, and optional subagents for git-spice stacked-branch workflows.
-
Receive encrypted Ola Ink Supernote notes in Pi
-
Issue a one-off model request from a pi extension the same way Pi core does: resolve auth, add provider session headers, call the completion
-
Parley is a durable conversation layer for Pi sessions: targeted messaging, explicit asks and replies, honest delivery receipts, and opt-in broker federation for remote peers — whether you're driving the conversation or letting agents coordinate. Local se
-
WebSearch tool extension for pi-agent.
-
Microsandbox-backed isolation for Pi coding tools
-
Cross-harness session hub for Pi: browse, search and continue sessions from Claude Code, Codex, OpenCode, Crush and JCode without leaving your chat.
-
Shows the live output token rate (tok/s) and time to first token (TTFT) of the current model in the pi TUI status bar.
-
Local PII redaction for Pi powered by cloakrs
-
Independent Git workspaces and session navigation for Pi, backed by Rift.
-
Pi extension: delegate git operations (diff/log/blame) to cheaper models via subagents, saving costs and keeping parent context clean.
-
Universal, configurable Codex-style apply_patch extension for Pi — works seamlessly with Claude, Gemini, DeepSeek, GPT, and custom API relays.
-
Auto mode for Pi: deterministic rules first, then a local decision model, then a confirm prompt
-
Temporary package placeholder for staged publishing
-
Shared ESLint, Prettier, Stylelint, Vale, Vite, and Vitest configs, custom lint rules, and pi coding-agent extensions - for keeping codebases maintainable when coding agents are the primary contributors.
-
Pi extension for monitoring Z.ai API token usage quota
-
Agent Memory Repo for pi — git-backed cross-session memory: memory_init/search/read/save tools, [[link]] navigation, auto-committed entries, session-start injection
-
Pick code blocks or inline snippets from recent assistant messages to copy or insert.
-
Quote text from assistant responses in Pi's active session branch
-
Resume a Pi session automatically after a rate limit: reads the reset hint, waits, then sends a marked continuation. Also waits out 5-hour and weekly plan windows by their real end time, and retries dropped streams and timeless failures on a backoff.
-
Aggregated turn tokens per second for Pi, excluding tools and gaps between model calls.
-
Curated Agent Skills (SKILL.md): structured thinking lenses, decision support, and brainstorming workflows for AI coding agents.
-
Auto-generate names for your pi session and the terminal surfaces it runs in (tmux window, herdr pane/tab, zellij pane/tab) so it is easy to see what each pi is doing. Fork of @normful/pi-auto-name with subagent-surface gating and config moved into pi's s
-
Minimal-intrusion task list extension for the Pi coding agent: a `todo` tool, a `/todos` command, and a live panel above the editor. State is replayed from the session branch, so it survives /reload and compaction.
-
Proactive AnySearch web access for Pi: one anysearch entry for 1-5 general or vertical queries with in-tool domain discovery, plus anysearch_extract for full pages. Built for current facts, source-backed claims, news, prices, papers, docs, and recommendat
-
NERVous System — durable multi-agent orchestration, coding-agent coordination, and workflow recovery for pi.
-
Switch Pi's AI provider and model based on time of day via YAML time-slot configuration.
-
Real-time debugger for Pi agent lifecycle events — streams turns, messages, tools, and provider events to a local webapp.
-
A read-only Pi command for inspecting registered tools, exposure, active status, and sources.
-
bridge any raw llm stream to be used with pi
-
Pi extension: renders provider/API errors as a readable, highlighted block with ctrl+o raw-data expansion instead of raw JSON.
-
Pi coding agent extension: /consult audits the current session with a second "auditor" model, then restores your original model.
-
Optional pi-chef recipe: the chef and every cook work by the pstack principles.
-
💎 LaPis — Persistent memory for the Pi coding agent. One SQLite DB, zero cloud, zero API keys.
-
Session counters in pi's footer: tool calls blocked, anchors injected, bytes kept out of context, actions queued by focus mode.
-
Previews syntax-highlighted fenced snippets and copies them individually from Pi assistant responses.
-
Pi extension that adds a /theme command for switching themes.
-
Native Codex compaction for Pi, directly or through CLIProxyAPI.
-
Show AI provider quota and balance in the pi TUI status widget
-
Local, read-only Pi session and system-prompt inspector