Windows · macOS · Linux

Your desktop
AI workspace.

Code, consult your sources, keep notes, and schedule follow-ups in GoodBuddy. Built-in integrations bring three coding agents to the desktop: OpenCode, Continue, and DeepSeek Harness (preview), alongside direct model chat.

  • Three desktop platforms, x64 and arm64
  • Built-in coding agents, configurable models
  • No GoodBuddy account required

Everyday workflows

From research to follow-up

Search project material, save decisions, and set reminders for what comes next.

Knowledge and citations

Find an answer and check its source

Ask questions about local documents or your organization's knowledge base. Follow citations in the answer to the original material.

Knowledge Q&A

What should we check before a staging release?

Check configuration, run regression tests, and confirm a rollback package is available. [1]

[1] Release handbook · Pre-release checks

Notes and schedules

Keep notes and schedule the next step

Save decisions in Magic Notes and schedule a reminder for work that needs a follow-up.

Project follow-up

Save these notes and remind me to check release readiness next Monday.

Saved: review the search fix and check regression results before release.

  • To-do: review the search fix
  • Schedule: next Monday, 09:00, release readiness check
Magic Notes · Release preparation

Product highlights

Projects, knowledge, and agents on your desktop

Choose an agent for the task, connect project files and knowledge sources, and review tool calls and results.

01

Built-in OpenCode, Continue, and DeepSeek Harness

Select a coding agent, configure its model, and start a task in your project from the desktop interface. DeepSeek Harness is in preview. Direct model chat is also available.

Direct modelsOpenCodeContinueDeepSeek Harness (preview)
02

Local and SSH projects

Work with local projects, or use OpenCode and Continue with remote projects over SSH. Remote projects require the companion gbagent on the remote host.

03

Connect existing knowledge services

Use a local knowledge base or configure Dify, FastGPT, and RAGFlow connections to retrieve material for your conversations.

04

Ask questions or give it a task

Available tools depend on the selected Runtime and enabled capabilities. Manage MCP services and their Runtime assignments in settings. Operations use the current account's permissions, and tool calls appear in the run history.

05

Skills and MCP tools

Import Skills and connect MCP services to give your chosen agent the tools you use at work.

06

Run history and token usage

Review tool calls, cancellation, timeouts, token usage, and run history in one place.

Desktop assistant

Conversations, knowledge, notes, and tasks on your desktop

Keep reference material, to-dos, and ongoing work in one desktop workspace. Choose project files for each conversation. See the configuration guide for each agent's context and tool support.

Choose your desktop release
01

Search files and web pages

Import files, folders, and web pages, then configure full-text and vector search. Enable the knowledge graph to explore relationships between sources.

02

Notes, to-dos, and follow-up

Magic Notes captures ideas and tasks. Heartbeat reviews work at configured intervals; both Heartbeat and scheduled work require the desktop app to be running.

03

Understand what is on your desktop

Add files, screenshots, app windows, clipboard content, and offline voice input when needed.

04

Reach your desktop assistant through messages

Configure messaging channels to send tasks in separate conversations. Message processing requires your desktop to be online and the channel connection to be available.

Before you start

Frequently asked questions

See the README getting-started guide for installation and connection steps.

Do I need an account or a paid model subscription?

No GoodBuddy account is required. External model services may need an account or API key and charge at their own rates. Local models need a suitable runtime on your machine.

Where does my data go?

Conversation history and notes are stored on your machine. Using models, knowledge services, MCP tools, or messaging channels sends relevant content to the services you configure. Model requests may include retrieved passages and attachments.

How do I set up a coding agent?

Select an agent, configure its model connection, and open a project. Prepare or download the runtime components required by your agent and operating system. DeepSeek Harness is in preview. See the getting-started guide for installation steps.

Which platforms and Chinese domestic systems are supported?

Windows, macOS, and Linux releases cover x64 and ARM64, including corresponding Chinese domestic hardware. LoongArch has a separate experimental preview. For systems such as UOS and Kylin, choose the matching architecture and package; compatibility information does not imply vendor or whole-system certification.

What do SSH projects require on the remote host?

Remote projects with OpenCode and Continue require a working SSH connection, a companion gbagent matching your desktop version, and the required runtime environment on the host. Operations use the remote login account's permissions. See the README for setup.