cristicretu/diri

The best way to work with coding agents. Run fifty in parallel, let them spawn each other, review everything in one place.

RustAGPL-3.0 Dernier commit il y a 2 jours diri.sh

Ouvrir sur GitHub

24 dernières heures: +6; étoiles
7 derniers jours: +19; étoiles
30 derniers jours: +131; étoiles
Étoiles au total: 438; 307 il y a un mois
Forks: 33; 13 contributeurs

Historique des étoiles

12 dernières semaines · étoiles cumulées par semaine · créé le 4 août 2026, zéro avant · calculé à partir des étoiles publiques de GitHub

Rang dans ses thèmes

Position parmi les dépôts classés dans chaque thème (min. 20 étoiles gagnées)

Rang de cristicretu/diri dans chaque thème pour aujourd’hui, cette semaine et ce mois-ci
Thème	Aujourd’hui	Cette semaine	Ce mois-ci
Rust	–	–	#168sur 407
Claude Code	–	–	#280sur 447
Agents de code	–	–	#310sur 529

Signaux de santé

En bonne santéVérifié le 14 oct. 2026

Activité des commits: Des commits chaque semaine depuis sa création (11 semaines) 674 commits sur la branche principale en 90 jours
Mainteneurs: 5 mainteneurs actifs Auteur principal : 89 % des 100 derniers commits
Réactivité sur les issues: Première réponse médiane : 13 heures Fermées : 35 sur les 50 dernières issues ouvertes en 90 jours
Pull requests: 40 fusionnées, 5 fermées, 5 ouvertes Sur les 50 dernières ouvertes en 90 jours
Versions: Dernière version : il y a 4 jours 39 versions ces 12 derniers mois
Hygiène du projet: Licence, CI et politique de sécurité
OpenSSF Scorecard: Non analysé Inconnu, pas compté contre le projet

Ces signaux sont calculés automatiquement à partir des données publiques de GitHub. Ce n’est ni une recommandation ni un audit de sécurité. Comment la santé est calculée

README

README.md

diri

The best way to work with coding agents.

Run five agents or fifty. Claude Code, Codex, Cursor, Gemini, and any other terminal agent, side by side. diri shows you which ones need you, gives each task its own worktree, lets agents spawn and coordinate other agents, and puts every change in one place to review. Native, built with Rust and GPUI. macOS, with Linux in beta.

Download · Getting started · Guides · Documentation · Contributing

Install

brew install --cask cristicretu/diri/diri

macOS 15 or newer. Apple silicon and Intel. Signed and notarized. You can also download the DMG and drag Diri to Applications.

Linux beta: x86_64 Ubuntu 22.04 / 24.04, X11 or Wayland, Vulkan 1.3. See the Linux guide for packages, source builds, and limitations. Linux packages are not included in every release.

Bring your own agents. diri runs the CLIs and accounts already on your machine, 22 of them out of the box. Claude Code and Codex get the deepest status and resume integration.

Run agents in parallel

Know what needs you. Live status and notifications separate working, waiting, and finished sessions. Read one sidebar, not thirty terminals.
Let agents run agents. The built-in MCP server lets an agent spawn other agents, hand them tasks, read their output, and answer their questions. A swarm is one prompt away.
Keep tasks separate. Every agent can get its own Git worktree and branch, so parallel work never collides.
Review in context. Inspect diffs, stage, commit, and follow pull request checks beside the session that wrote the code.
Never lose a session. Each terminal is owned by its own process. Close the app or restart the Engine and every agent keeps running.
Use your own machines. Run locally or on any SSH host you control. No tmux, no sudo, no Diri account, no hosted relay.

Where this is going

Agents are good enough now that the bottleneck is you: how many you can keep track of and how much context you can hold. diri exists to remove that.

Notes are next. Write a PRD in a note, break it into to-dos, and send any to-do to an agent. Agents write notes back. You stay at the level of the plan while the agents do the work.

Today you can run dozens of agents in parallel. The goal is hundreds.

Contribute

Small, well-tested changes are welcome. Start with a reproducible bug, a focused fix, clearer documentation, or an agent manifest. Read the contributor guide for setup and review expectations.

Report a bug · Discuss an idea · Roadmap