Tobub: Desktop Browser Automation for AI Agents

A desktop app that lets AI agents run reusable website workflows in your local browser and saves the results for review.

Website · Download Tobub · Skill examples · 中文-1.8.5

What is Tobub?

Tobub is a desktop app for automating website tasks with AI agents. You can run a Skill yourself or let an agent choose one; Tobub operates the browser on your computer and saves each run and its results in a Session—a work record you can review or continue later.

A Skill is a saved, reusable workflow for one task, such as collecting recent Reddit posts or YouTube comments.

This repository contains Skill definitions as JSON files, examples, separate Agent Skills for Claude Code and local Codex, a WorkBuddy Connector, and a DeepSeek Harness plugin. The desktop app is distributed separately; you need it to run these workflows.

Product demo

Why use Tobub for browser automation?

	What it means
Digital Employees	AI Agents with specialized roles that plan work, coordinate Skills, and carry multi-step tasks forward.
Automation Skills	Reusable JSON workflows that connect Agents to supported platforms and tools for research, analysis, and action.
Work Sessions	Persistent records of plans, sources, progress, results, and context, ready for review or follow-up.

flowchart LR
    A[Your goal] --> B[Digital Employee]
    B --> C[Plan the work]
    C --> D[Compose Skills]
    D --> E[Run through Tobub Runtime]
    E --> F[Useful result]
    F --> G[Saved Session]
    G --> B

How automation Skills work

A Skill is a saved workflow that tells Tobub how to carry out a task on a supported website—for example, collecting Reddit posts or YouTube comments. A typical Skill works in four stages:

1. Collect information from supported websites and services.
2. Organize results into structured, reusable work material.
3. Analyze trends, customer activity, competitors, and market signals.
4. Execute the requested next step and return a clear result.

Skills become more useful when combined. For example, a Research Analyst can connect Search, Web Crawler, RSS, GitHub, and community Skills into one repeatable market-intelligence workflow.

Digital employee roles

Tobub currently presents these Agent roles:

• Operations Specialist
• Data Engineer
• Research Analyst
• Content Strategist
• Social Media Manager
• Market Intelligence Analyst
• SEO Specialist
• Customer Insights Analyst
• Product Analyst
• News Desk Editor
• Sales Researcher
• Knowledge Curator

Each role turns a professional method into a reusable workflow. The Agent coordinates Skills, keeps context, tracks progress, and leaves work that people can review and continue.

Current automation Skills

• Research: Google Search, Web Crawler, RSS & Feeds, Reddit
• Social intelligence: YouTube, Facebook, TikTok, Instagram, X / Twitter
• Business and market: LinkedIn, Shopify / Amazon
• Developer intelligence: GitHub

Use these Skills for market research, trend discovery, social listening, customer insight, competitor monitoring, news monitoring, SEO research, and structured web extraction.

Typical workflow

1. Give Tobub a concrete outcome.
2. The digital employee confirms the goal and creates a step-by-step work plan.
3. It selects the Skills needed for the task.
4. Skills run through the Tobub Runtime against supported platforms.
5. Tobub keeps progress, sources, supporting evidence, and intermediate results visible in the Session.
6. The Agent delivers the result and saves a Session for follow-up work.

Use Tobub with AI platforms

Each integration connects a local agent to the Tobub desktop app. Start Tobub on the same computer, then follow the guide for your platform:

• Claude Code guide — Agent Skill + MCP.
• Codex guide — Agent Skill + MCP.
• DeepSeek Harness guide — native plugin.
• WorkBuddy guide — Connector.

Open-source Skills for Tobub. Runtime distributed separately.

This repository publishes Tobub’s open Skill layer: reusable JSON definitions that compatible Tobub software can install and execute. A Skill is an automation workflow, not a standalone browser automation application.

The Tobub desktop application and Runtime implementation are proprietary products distributed separately. Cloning this repository alone does not provide the app, platform adapters, authentication, account management, or complete execution environment.

Open in this repository	Not included here
Native JSON Skill definitions	Tobub desktop application
Skill documentation and examples	Tobub Runtime implementation
Claude Code and Codex Agent Skills, WorkBuddy Connector, and DeepSeek Harness plugin source	Platform adapters and authentication
Community review and improvements	Desktop installers and commercial distribution

This separation lets the community review and extend automation workflows while Tobub supplies the local execution environment and product experience.

Running a Skill

To run a Skill in Tobub:

1. Install Tobub for Windows, macOS, or Linux.
2. Import or install a Skill from this repository.
3. Configure the platform account and runtime resources required by that Skill.
4. Run it directly or let a digital employee select it.
5. Review the result and execution history in the Session.

The Tobub application also supports Docker and self-hosted deployment, local data control, and OpenAPI integration. These product capabilities are provided by the separately distributed application and Runtime.

Contributing a Skill

Make every Skill contribution easy to understand, review, and run safely:

• explain the problem and workflow it improves;
• state the platform, inputs, outputs, side effects, and required permissions;
• include the JSON definition and a realistic example;
• document rate limits, platform restrictions, and failure behavior;
• test it inside compatible Tobub software before opening a pull request;
• never commit credentials, cookies, access tokens, private data, or unauthorized content.

Maintainers should install the local pre-commit check once with bash scripts/install-git-hooks.sh. It verifies the staged demo and Claude Code/Codex Agent Skill files before commit. If a demo changes, rebuild both generated packages from the SDK using the commands in the platform package maintenance guides, then stage the generated files.

Compatibility and safety

Skills interact with third-party platforms. Their terms, APIs, rate limits, privacy requirements, robots rules, and local laws still apply. Use test accounts and minimum permissions when validating a Skill. A Skill should fail closed when authentication, required input, or platform state is uncertain.

Product access

• Download Tobub
• Tobub website
• 中文 README
• MIT

The desktop application is distributed separately from this open Skill repository.