Danmo Work

Danmo Work

[English](README.md) | [中文](README.zh-CN.md)

[](https://github.com/danmo-ai/danmo-work/releases/latest) [](LICENSE) [](go.mod) [](https://github.com/danmo-ai/danmo-work)

Open-source AI work agent — multi-agent, long-horizon work on your own machine. Coding, research, reports, slides, sheets, and automations all run on a single agent loop you can watch, correct, and resume. Self-hosted, MIT.

• The model orchestrates. No workflow graphs to configure. Subagents (delegate_agent), questions (ask_user), and MCP connectors are all just tools the agent calls as it plans.
• Turn Log is state. Every tool call is appended to JSONL. Recover from a crash, replay a turn, or edit a tool result and let the loop continue.
• Everywhere. Web · Desktop (Tauri) · CLI · TUI · WeChat / Feishu / WeCom / QQ — one engine, tools always running on your machine.

[Interactive tour](docs/demo/product-tour.html) · [MP4](docs/demo/product-tour-en.mp4) · [Office co-edit tour](docs/demo/office-coedit-tour.html?lang=en&tour=1)

> Code, reports, decks, sheets, demos, automations — on one trail.

---

Highlights

| | | |--|--| | Multi-agent teams | A lead agent summons built-in specialists — Document, Implementer, Researcher, Reviewer, GitHub, Novel Writing… — each in its own isolated context | | Document Stage + AI Diff | Docs, slides, and sheets on one canvas: propose → review diff → keep / revert / accept hunks. Markdown stays the source of truth | | Humans in the loop | ask_user is just another tool — with options and forms. Approval gates stop risky commands before they run | | Plans are a tool too | The plan is todowrite: the agent updates status as it works, marks items done only after verifying, and rewrites the plan when it drifts — not a one-time artifact that gathers dust | | Durable memory & Table Store | memory_update / memory_read across user / project / agent scopes; table_* schema-free business rows; Markdown knowledge bases with chapter-level search | | Skills & connectors | Skill market (official catalog, Tech Leads Club, ClawHub), disk-scanned custom skills, MCP connectors bound per agent | | Real sandbox | OS-level shell sandbox (Seatbelt / Landlock / bwrap / WSL2), network deny / domain allowlists, four permission modes | | Automations | Cron schedules and webhooks start agent turns in the background | | Your model, your data | Anthropic-native + OpenAI-compatible providers (OpenAI, DeepSeek, GLM, Qwen, Kimi, Gemini, Grok, local Ollama…). Data lives in ~/.danmo-work/ |

---

Design

Three ideas run through the whole system (full architecture: [docs/core-design.md](docs/core-design.md)):

1. Everything is a tool. Files, shell, web, memory, tables, knowledge, subagents (delegate_agent), even humans (ask_user) share one interface. The model decides what to call and when.
2. The model orchestrates. No developer-written control flow: the lead agent plans, delegates, and checks in with you on its own. Code and long-running work are the same loop at different depths — there is no mode switch.
3. Logs are state. Each turn is an append-only JSONL trail of tool calls. Sessions can span days or weeks; recovery is built in, not bolted on.

Experts & teams

The lead agent runs the session; specialists do focused work on demand, each in an isolated sub-turn (roster and usage: [docs/experts.md](docs/experts.md)):

| Agent | Role | |-------|------| | Team (lead) | Multi-agent by default; collaboration toggle | | Document | Reports, slides, sheets (Markdown source for Document Stage) | | Comms | Polishes messages, emails, notifications | | Implementer | Code changes from specs (TDD / debugging skills) | | Explorer | Read-only codebase exploration | | Researcher | Deep research and retrieval | | Reviewer | Code and artifact review | | Data | CSV / JSON analysis and reporting | | GitHub | Issues, PRs, Actions, releases | | Danmo Make | Local image / video / audio generation (separate app) | | Novel Writing | Long-form fiction: outline → chapter contract → draft → review → commit | | CodeGraph (market) | Code intelligence — definitions, references, impact |

Summon a specialist with @ in the Composer, or just ask in plain language ("delegate the Document expert…"). The lead only sees the subagent's report, so the main context stays lean and KV-cache friendly. You can also build your own subagents in Teams, binding skills, tools, knowledge bases, and connectors.

Beyond experts, the same library hosts skills (one-click workflows — document-writing, playable-slides, TDD, deep-research…) and connectors (MCP integrations). Install from the market, bind them per agent, or drop your own skills into ~/.danmo-work/skills/.

Document Stage

Three panes — projects · agent stream · right panel (Plan / Files / Memory / Changes / Terminal) — around a central Document Stage whose toolbar follows the file kind (doc / slides / sheet / code / diff / preview). In Preview you can click a DOM element, annotate it, and send the exact HTML/CSS context to the Composer.

Office files are co-edited as a normal agent turn with a review step:

1. Intent — select text, a slide, or cells, type an instruction → [office-edit]
2. Propose — the agent edits; a snapshot is taken before the turn
3. Review — view the diff; keep, revert, or accept individual hunks
4. Commit — keep persists, revert restores the snapshot; the trail stays in the Turn Log

Safety

• Soft gate — permission modes (discuss / plan / interactive / auto); risky commands ask before running
• Hard sandbox — OS-level isolation for shell, network deny or domain allowlists
• auto_approve never silently approves dangerous commands; approvals render inside IM chats too

Channels, remote & automations

• IM on the same loop — Feishu, QQ, WeChat, and WeCom connect outbound to your own accounts; progress cards, questions, and approvals appear in-channel. Tools still run on your machine.
• Remote Hub — pair this PC with danmo-hub so remote clients can drive it, even behind NAT.
• Automations — cron and webhook triggers run sessions while you're away.

---

Install

| Platform | Package | |----------|---------| | macOS (Apple Silicon) | Homebrew or .dmg | | Windows | Setup .exe | | Linux (x86_64) | AppImage / .deb |

All binaries: GitHub Releases.

macOS

brew tap danmo-ai/tap
brew install --cask danmo-work
# upgrade: brew update && brew upgrade --cask danmo-work

Fallback: brew tap danmo-ai/danmo-work https://github.com/danmo-ai/danmo-work.git Or download Danmo.Work_*_arm64.dmg from Releases. Not Apple-notarized yet — first launch: right-click → Open.

Windows / Linux

• Windows: Danmo.Work_*_x64-setup.exe — until Authenticode is enabled, SmartScreen may warn → More info → Run anyway.
• Linux: chmod +x Danmo.Work_*_amd64.AppImage && ./… or sudo apt install ./Danmo.Work_*_amd64.deb (needs WebKitGTK).

First run

Open the app, add an LLM API key in the UI (or ~/.danmo-work/config.yaml), and pick a model. Projects, sessions, memories, and the rest are created for you under ~/.danmo-work/.

From source

Needs sibling dq-ui. Then:

make dev-web   # backend :7801 + Vite :5801 → http://localhost:5801/app/

---

Development

Prerequisites: Go 1.26+, Node.js 20+, sibling dq-ui.

make dev-web          # backend :7801 + Vite :5801 → http://localhost:5801/app/
make stop             # stop all DQ_DEV processes
mkdir -p ~/.danmo-work && cp config.example.yaml ~/.danmo-work/config.yaml

Build, pack, test, env vars, and CI: [AGENTS.md](AGENTS.md). Architecture: [docs/core-design.md](docs/core-design.md).

---

Docs

| Doc | Description | |-----|-------------| | [docs/core-design.md](docs/core-design.md) | Agent architecture, tool system, channels, Document Stage | | [docs/experts.md](docs/experts.md) | Expert usage and the built-in roster | | [docs/remote/README.md](docs/remote/README.md) | Remote Hub pairing and tunnel protocol | | [docs/demo/README.md](docs/demo/README.md) | Product tour HTML / GIF / MP4 | | [evals/dqharbor/README.md](evals/dqharbor/README.md) | Terminal-Bench 2.0 eval harness | | [AGENTS.md](AGENTS.md) | Contributor quick reference | | [config.example.yaml](config.example.yaml) | Full config reference |

License

[MIT](LICENSE)

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: danmo-ai
• Source: danmo-ai/danmo-work
• License: MIT
• Homepage: https://github.com/danmo-ai/danmo-work/releases/latest

Install and usage instructions live in the source repository linked above.