diri

Name: diri
Author: cristicretu

Verified

Native macOS orchestrator for coding agents — run Claude Code, Codex, Cursor, Gemini and shells in parallel across git worktrees and remote hosts

258stars

23forks

Rust

Installation

# Add to your Claude Code skills
git clone https://github.com/cristicretu/diri

Getting Started

Guides for using ai agents skills like diri.

Caveman: Cut Claude Token Use by 65%
How agent-side prompt compression works, when to use it, and when not to.
What is an AI Skills Marketplace?
Definitions, how marketplaces work, and how to choose between them in 2026.
Getting Started with AI Skills

Security ReportVerified

Last scanned: 8/5/2026

{
  "issues": [],
  "status": "PASSED",
  "scannedAt": "2026-08-05T06:28:12.820Z",
  "npmAuditRan": true,
  "pipAuditRan": true,
  "promptInjectionRan": true
}

README.md

Frequently Asked Questions

What is diri?

diri is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by cristicretu. Native macOS orchestrator for coding agents — run Claude Code, Codex, Cursor, Gemini and shells in parallel across git worktrees and remote hosts. It has 258 GitHub stars.

Is diri safe to use?

Yes. diri passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.

How do I install diri?

Clone the repository with "git clone https://github.com/cristicretu/diri" and add it to your Claude Code skills directory (see the Installation section above).

What programming language is diri written in?

diri is primarily written in Rust. It is open-source under cristicretu on GitHub, so you can review or fork the full source.

Are there alternatives to diri?

Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh diri against similar tools.

Agentic AI for Beginners

Build your first AI agent from scratch - tool use, ReAct pattern, memory, deployment

41 minBeginner

Comments (0)

to leave a comment.

No comments yet. Be the first to share your thoughts!

Related Skills

ECC

by affaan-m

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

240,529

cc-statusline hass_local_openai_llm

Native desktop orchestrator for coding agents on macOS and Linux. Run Claude Code, Codex, Cursor, Gemini and plain shells in parallel — across git worktrees or on remote hosts — each with a live status (working / needs-you / done) and tmux-like persistence: closing the app never kills a session, and a daemon restart brings conversations back.

diri

Install

macOS

brew install --cask cristicretu/diri/diri

Or download the latest DMG from Releases, open it, and drag diri to Applications. Either way it is the same universal build (Apple silicon and Intel), signed and notarized. diri updates itself from there.

The tap has to be named in full — a bare diri resolves only against Homebrew's default taps. The cask lives in cristicretu/homebrew-diri rather than homebrew-cask, which requires a notability threshold diri does not meet yet.

macOS 15 or newer.

Linux beta

Download the x86_64 AppImage or Debian package from Releases. Ubuntu 22.04 and 24.04 are supported under X11 and Wayland with a Vulkan 1.3-capable GPU.

sudo apt install ./diri_<version>_amd64.deb
# or
chmod +x diri_<version>_amd64.AppImage && ./diri_<version>_amd64.AppImage

See the Linux beta guide for checksums, upgrade and uninstall steps, XDG paths, graphics troubleshooting, and current limitations.

60-second tour

Add a project directory and create a session for Claude Code, Codex, another supported agent, or a plain shell.
Start several sessions, ideally in separate git worktrees when they edit the same repository.
Watch the sidebar instead of every terminal: it shows which agents are working, waiting for you, or done.
Quit and reopen diri. The daemon keeps each PTY alive and replays the session when you return.

The getting-started guide covers remote hosts, MCP orchestration, diagnostics, local data, and uninstalling.

What it does

Many agents at once. Each session is a real terminal with a real PTY. Group them by project, split them across git worktrees, or run them on a remote host over ssh+tmux.
Status you can trust. diri reads what an agent actually painted on its screen and tells you which ones are working, which are waiting on you, and which are done — so you can watch ten sessions without reading ten terminals.
Sessions outlive the app. A background daemon owns the PTYs. Quit diri, reopen it, and everything is still there.
Agents can orchestrate agents. An MCP server lets a running agent spawn another one, watch it, read its output, and answer its prompts.

First-class status detection and resume are Claude Code and Codex. Cursor and Gemini run with partial support, and anything else runs as a terminal with running/exited status.

Claude Code, Codex, Cursor, Gemini, and shell agents

Architecture

Two processes, one wire protocol:

Diri architecture: app and CLI connect through the control socket to the engine, persistent PTY holders, and coding agents

diri — the desktop app: Rust + GPUI. Owns the window, sidebar, terminal renderer, command palette, and usage accounting. Lives in diri/.
dirijord-rs — the headless Rust engine, launched by the app and outliving it. Owns PTYs and child agent processes, an offset-addressed output log per session (for detach and replay), a headless terminal emulator for status detection, the session registry and persistence, worktrees, and the control socket.

dirijor is the automation CLI for hooks, notifications, status, and diagnostics; dirijor-mcp is the MCP stdio server injected into agents. diri-holder owns each PTY master so sessions survive an engine restart. Every shipped executable is built from the Rust workspace in diri/.

Adding an agent

Agent support is data, not code. Each agent is one JSON file in diri/crates/diri-engine/manifests/ describing how to spawn it, how to resume, which keys approve or deny a prompt, and the screen rules that decide whether it is working, waiting, or done. Copy the closest existing manifest and adjust it — no code changes required. The manifest-authoring guide explains the schema, safe capture workflow, examples, overrides, and validation. This is the easiest way to contribute.

Building from source

Needs Rust (pinned in diri/rust-toolchain.toml) plus the platform build dependencies. On macOS that means the Xcode command-line tools; Linux dependencies and packaging commands are listed in diri/LINUX.md. The first build compiles GPUI from a pinned Zed revision and takes a while.

(cd diri && cargo build)                   # app, engine, holder, CLI, MCP
(cd diri && cargo test --workspace)
(cd diri && cargo run -p diri-app)         # run the app from source

diri/scripts/package.sh                    # full bundle
diri/scripts/package-linux.sh              # AppImage + DEB, on x86_64 Linux
diri/scripts/install-local.sh

Run the same core checks as CI with one command:

./scripts/check.sh

diri/PACKAGING.md covers signing and notarization, diri/UPDATING.md the updater and release flow, diri/NODE.md running agents on a remote VPS node.

The documentation index links user and engineering guides.

Contributing

See CONTRIBUTING.md. Bug reports, fixes, docs, and new agent manifests are all welcome. New contributors can start with good first issue or help wanted.

Questions belong in Discussions, reproducible bugs in Issues, and vulnerabilities in private security reports. See the roadmap, support guide, privacy notice, and governance for project expectations.

License

Diri's original source is Apache 2.0. Builds also contain third-party software under its own licenses; see LICENSE, NOTICE, and the machine-checked dependency policy in license-policy.json.