Proqi
A terminal-native prompt composer for power users running multiple AI coding agents (Claude Code, Codex, and more) at once.
🔗 Visit Proqi
Description
If you're only running one AI coding agent at a time, you probably just type into its chat box and move on. But once you're juggling several agents on different tasks — one refactoring a module, another writing tests, a third exploring a bug — keeping track of which prompt goes where, and being able to edit and reuse them, turns into its own small management problem. Proqi is built specifically for that stage: a dedicated, keyboard-driven app for composing, organizing, and dispatching prompts to multiple agents.
Proqi is a Rust application (using the Ratatui terminal UI library and a local SQLite database) that lets you write prompts as independent, editable “thoughts” organized on boards, reorder and duplicate them, and submit them to multiple coding agents at once through an integration called Herdr. It supports Claude Code, Codex, Pi, Hermes, and conditionally OpenCode and Kilo, includes a macOS-only screenshot inbox with annotation for visual prompts, auto-saves everything locally with undo/redo, and ships with zero telemetry or cloud sync — it's MIT-licensed and installs via Homebrew or Cargo.
💬 Our review
The short version: if you're already running several AI coding agents in parallel and losing track of which prompt went where, Proqi's board-based prompt management is a genuinely useful missing piece — but if you only run one agent at a time, this solves a problem you don't have yet.
There's no direct named competitor for exactly this niche (multi-agent prompt composition), because most people either type directly into each agent's own interface or use plain text files as scratch notes — Proqi's value is formalizing that scratch-note habit into something searchable, reorderable, and dispatchable to multiple agents via its Herdr integration. Being local-first with zero telemetry and a real SQLite-backed undo/redo system is a meaningful trust signal for a tool that's going to hold your work-in-progress prompts. The screenshot annotation feature is macOS-only, which will disappoint Linux users, and this is still a young, single-purpose tool so expect some rough edges. Worth adopting now if you're already deep into running multiple coding agents in parallel; skip it if you're a single-agent, single-task workflow.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit et open source (MIT).
Pros
Pensé spécifiquement pour la gestion de prompts multi-agents
Local-first, zéro télémétrie
Organisation en boards avec undo/redo et sauvegarde SQLite locale
Gratuit, open source, rapide (Rust)
Cons
Ne résout un problème que si vous faites déjà tourner plusieurs agents en parallèle
Annotation de captures d'écran limitée à macOS
Projet jeune, adoption encore limitée