Coding & Developmentπ€claudeπuniversal
AgentDesk
AI agent orchestration platform β single binary, hot-reloadable policies
Quick Start
Install this Skill
npx skills add itismyfield/AgentDesk# AgentDesk > AI agent orchestration platform β a single Rust binary that manages teams of AI agents through Discord, with a web dashboard, kanban pipeline, and hot-reloadable policy engine. AgentDesk lets you run multiple AI agents (Claude Code, Codex, or any CLI-based provider) as a coordinated team. Agents communicate through Discord, execute tasks via tmux sessions, and follow configurable workflows defined in JavaScript policy files. ## Quick Start ### One-Click Install (macOS) ```bash curl -fsSL https://raw.githubusercontent.com/itismyfield/AgentDesk/main/scripts/install.sh | bash ``` This will: 1. Download the latest release (or build from source if no release is available) 2. Install to `~/.adk/release/` 3. Create `~/.adk/release/config/agentdesk.yaml` 4. Register a launchd service (auto-starts on boot) 5. Open the web dashboard for guided onboarding Set `AGENTDESK_INSTALL_REPO` or `AGENTDESK_INSTALL_DIR` before running the installer when you need to test a fork or install into a sandboxed runtime root. Non-default install roots get an isolated launchd label and a derived non-8791 loopback port automatically; set `AGENTDESK_LAUNCHD_LABEL` or `AGENTDESK_INSTALL_PORT` only when you need explicit values. The default install root keeps the existing `com.agentdesk.release` label path so the bootstrap remains compatible with the current published release binary. For a portable operator scaffold before wiring real Discord or connector secrets, run: ```bash python3 scripts/operator-init-portable.py --root ~/.adk/release ``` The helper writes `config/agentdesk.yaml`, prompt stubs, workspace directories, and `config/launchd.env.example` without embedding machine-specific `/Users/<name>` paths in repository templates. Use `--with-obsidian-stubs` only when you want local placeholder directories for the optional Obsidian connector. ### Windows and Linux Native Runtime Windows and Linux run natively today, but they use the manual/runtime-first path instead of the macOS `curl | bash` bootstrap. 1. Download the matching release artifact (`agentdesk-linux-<arch>.tar.gz` or `agentdesk-windows-<arch>.zip`) or build from source with `cargo build --release`. 2. Run `agentdesk init` (`agentdesk.exe init` on Windows) to create the runtime under `~/.adk/release` or `%USERPROFILE%\.adk\release`. 3. Start `agentdesk dcserver` directly, or register the generated service path with `systemd --user` on Linux or `nssm` / `sc.exe` on Windows. 4. Verify the host runtime with `agentdesk doctor`. Mutating repairs require explicit flags such as `agentdesk doctor --fix --allow-restart` or `agentdesk doctor --fix --repair-sqlite-cache`. When tmux is unavailable, provider turns automatically fall back to `ProcessBackend` instead of tmux sessions. That path still posts ADK session heartbeats every 30 seconds, but live child processes cannot be reattached after a `dcserver` restart. After restart, AgentDesk starts a fresh backend process on the next turn and only restores provider-native session IDs when the underlying CLI supports resume. ### Build from Source ```bash git clone https://github.com/itismyfield/AgentDesk.git cd AgentDesk cargo build --release # Verify the dashboard with the same command CI uses (Node >=22) ./scripts/verify-dashboard.sh # Initialize ./target/release/agentdesk init ``` #### Shared rustc Cache with `sccache` AgentDesk intentionally keeps a separate `target/` directory per worktree. Sharing `CARGO_TARGET_DIR` across always-parallel worktrees causes Cargo lock contention, so the supported acceleration path is a shared `sccache` rustc cache instead. - `.cargo/config.toml` no longer hard-codes `rustc-wrapper`, so bare `cargo build` / `cargo test` / `cargo check` works on every machine regardless of whether `sccache` is installed. No more `RUSTC_WRAPPER=` workaround for agents or subagents. - for bare Cargo commands, opt in manually with `export RUSTC_WRAPPER=sccache` after installing `sccache`. This uses inherited settings or sccache's own defaults; it does not apply AgentDesk's cache size or idle settings. - release/deploy scripts and the installer's source-build path call `setup_sccache_env` from `scripts/_defaults.sh` to activate caching when sccache is installed. - Linux/Windows CI configures sccache through workflow environment settings and the Mozilla action. macOS CI runs only on GitHub-hosted runners and disables sccache. - the shell helper and POSIX campaign wrapper (`scripts/build_token.py`) default to `SCCACHE_CACHE_SIZE=40G` and `SCCACHE_IDLE_TIMEOUT=0` (no idle exit). Unset or empty values use these defaults; nonempty overrides remain intact. The campaign wrapper skips activation entirely when `RUSTC_WRAPPER` or `CARGO_BUILD_RUSTC_WRAPPER` is already present, even empty, so a manually configured wrapper does not receive these defaults. The 40G limit allows up to 30G more local disk use than the previous limit; it does not preallocate that space. New environment settings do not rec ... (truncated)
Repository
9stars
5forks
Details
Related Skills
More coding & development skills you might like
Reticle
Coding & DevelopmentAI agents can generate code, but still struggle to understand what they build. Reticle brings Jev-style machine-native runtime perception to web & desktop applications.
π€claudeπuniversal
1.2k
190
View
agentscope-java
Coding & DevelopmentExpert Java developer skill for AgentScope Java framework - a reactive, message-driven multi-agent system built on Project Reactor. Use when working with reactive programming, LLM integration, agent orchestration, multi-agent systems, or when the user mentions AgentScope, ReActAgent, Mono/Flux, Project Reactor, or Java agent development. Specializes in non-blocking code, tool integration, hooks, pipelines, and production-ready agent applications.
πuniversal
6.0k
1.5k
View
ospec
Coding & DevelopmentDocument-driven OSpec workflow for initialization, change/goal routing, validation, archiving, and durable project knowledge.
β¨geminiπ€claudeπ¬gptπuniversal
524
34
View
misakanet-failure-memory
Coding & Developmentπ A zero-dependency, git-backed micro-lesson library for AI Agents to asynchronously share and search verified debugging experience. Python stdlib only. | https://misakanet.org
π€claudeβ¨οΈcursorπuniversal
424
158
View