Skip to main content
LOCAL & REMOTE AGENT TOOLS

Let agentsfinish real work

Connect an MCP-compatible AI client to your computers and servers, then use files, commands, Git, browsers, and external services through explicit tool boundaries.

Choose an installationExplore the features
30built-in tools
Recoverabletask state
Cross-platformlocal and remote
AgentDock Runtime connected
ACTIVE TASKUpdate the project and verify the result
3 / 4
✓Inspect the real environmentread_file
✓Apply the code changesfile_edit
✓Run tests and buildsexec_command
Publish and verify the resultbrowser_snapshot
Task state persistedResume at any time →
WHAT YOU CAN DO

From installation to real work

Start with the shortest installation path, then enable Skills, browser automation, external MCP servers, and long-running tasks as needed.

START

Install and connect

Follow the macOS, Windows, Linux, or Docker guide, then add the MCP URL to your client.

WORK

Work with files and code

Let the agent read projects, run commands, edit files, and operate Git, then verify the real result.

SKILLS

Use Skills

Install trusted Skills so agents can follow proven workflows, dependencies, and safety boundaries.

BROWSER

Operate a browser

Open pages, click, type, and capture screenshots while checking page, console, and network errors.

MCP

Connect external services

Connect additional MCP servers on demand and keep credentials in isolated environments instead of registry data.

TASKS

Track complex tasks

Persist goals, steps, progress, and verification so interrupted work can continue without mistaking execution for completion.

GET STARTED

Install for your current system

Most users should choose the native macOS, Windows, or Linux installer. Choose the container option when Docker is already part of your environment.

Choose an installationBrowse all documentation