Skip to content

ceasarXuu/AgentShare

BranchesTags

Repository files navigation

AgentShare - Multi-Agent System

中文文档

AgentShare Deployment Tool

AgentShare Intelligent Deployment Tool - Supports cross-platform Agent/Skill management with a modern TUI experience using Node.js + Ink.

Features

  • 🔍 Platform Detection: Detect installed platforms from schema-defined paths
  • 📊 Overview Dashboard: Show global/project Agent & Skill counts per platform
  • 🤖 Agent Management: Deploy/Extract/Uninstall Agents (where supported)
  • 🛠️ Skill Management: Deploy/Extract/Uninstall Skills (where supported)
  • 🔧 MCP Sync: Apply MCP server config when deploying Agents (where supported)
  • 🎨 Modern UI: Smooth interactive interface using Ink
  • 🌐 Multi-language: Built-in English and Simplified Chinese support (Startup Selection)

Quick Start

Method 1: Global Installation (Recommended)

Run the install script to add agentshare command to system PATH:

./install-macos-linux.sh

After installation, you can use it anywhere:

agentshare

Method 2: Windows Installation (PowerShell)

Ensure Git (Git Bash) is installed. Run in PowerShell:

.\install-windows.ps1

This will create agentshare command and configure necessary environment.

Method 3: Direct Run

If you don't want global installation:

./AgentShare.sh

Install Script Features

install-macos-linux.sh automatically performs:

  1. ✅ Detect OS (macOS/Linux)
  2. ✅ Check Node.js environment
  3. ✅ Verify AgentShare.sh exists
  4. ✅ Create global launcher
  5. ✅ Install to /usr/local/bin or ~/.local/bin
  6. ✅ Configure PATH environment variable
  7. ✅ Test installation

Uninstall

To uninstall, just delete the launcher:

# If installed in /usr/local/bin
sudo rm /usr/local/bin/agentshare

# If installed in ~/.local/bin
rm ~/.local/bin/agentshare

Main Functions

  1. Deploy Agent (Project → Platform)

    • Deploy repository Agents to platforms that support Agents (based on schemas)
    • Support simultaneous deployment to multiple platforms
    • Sync MCP server config when supported by the platform

  2. Extract Agent (Platform → Project)

    • Extract Agent definitions from a platform back into the repository

  3. Uninstall Agent

    • Remove an Agent from a platform (where supported)

  4. Skill Management

    • Deploy/Extract/Uninstall Skills to/from platforms that support Skills
    • Support both global scope and project scope (when platform schema provides project paths)

Available Agents

1. kuko

Business Opportunity Researcher Designed for independent developers and small teams to quickly identify the most valuable App/software product directions worth pursuing, and advance opportunities from "ideas" to "verifiable solutions" within limited time.

Project Structure

  • agents/: Agent definitions (neutral format)
    • <agent_name>/agent.md: Agent definition file (Markdown + frontmatter)
    • <agent_name>/docs/: Optional agent docs
    • <agent_name>/workflows/: Optional workflows
  • skills/: Reusable Skills (each skill has a SKILL.md)
  • schemas/: Platform schemas that define detection paths, outputs, and capabilities

Quick Start

Using kuko

  1. Install and run AgentShare: agentshare
  2. Select Agent Management → Deploy Agent and choose kuko
  3. Select the target platform and scope (global/project) where supported

API Key Configuration

Required API Keys

If you use MCP servers that require credentials, configure these environment variables where your platform reads them:

  1. GITHUB_TOKEN - For GitHub MCP tool
  2. BRAVE_API_KEY - For Brave Search MCP tool

Not Required Here

The following AI model API keys are typically managed by your platform (e.g., OpenCode/Claude Code) rather than this repository:

  • ❌ ANTHROPIC_API_KEY (Claude models)
  • ❌ OPENAI_API_KEY (GPT models)
  • ❌ GEMINI_API_KEY (Gemini models)

Why MCP Tool Keys Are Needed

MCP Tools vs AI Models

┌─────────────────────────────────────────────────────────┐
│                        kuko Agent                        │
├─────────────────────────────────────────────────────────┤
│                                                          │
│  AI Models (Managed by Platform)                        │
│  ├─ Claude (Anthropic)  ← OpenCode/Claude Code managed │
│  ├─ GPT (OpenAI)        ← OpenCode/Claude Code managed │
│  └─ Gemini (Google)     ← OpenCode/Claude Code managed │
│                                                          │
│  MCP Tools (Require Independent Configuration)          │
│  ├─ GitHub MCP          ← Requires GITHUB_TOKEN         │
│  ├─ Brave Search MCP    ← Requires BRAVE_API_KEY        │
│  ├─ Playwright MCP      ← No key required               │
│  ├─ Fetch MCP           ← No key required               │
│  ├─ SQLite MCP          ← No key required               │
│  └─ Filesystem MCP      ← No key required               │
│                                                          │
└─────────────────────────────────────────────────────────┘

Specific Use Cases

GitHub MCP (requires GITHUB_TOKEN):

  • Search open-source projects
  • Analyze competitor tech stacks
  • Review code implementations
  • Assess technical feasibility

Brave Search MCP (requires BRAVE_API_KEY):

  • Perform web searches
  • Collect market information
  • Find user discussions
  • Obtain trend data

How to Obtain API Keys

1. GitHub Token

Steps:

  1. Visit https://github.com/settings/tokens
  2. Click "Generate new token" → "Generate new token (classic)"
  3. Set name: kuko Agent
  4. Select permissions:
    • ✅ public_repo (access public repositories)
    • For private repos, select repo
  5. Click "Generate token"
  6. Copy the generated token (shown only once!)

Example:

ghp_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

2. Brave Search API Key

Steps:

  1. Visit https://brave.com/search/api/
  2. Click "Get Started" or "Sign Up"
  3. Create an account and log in
  4. Find your API Key in the Dashboard
  5. Copy the API Key

Example:

BSAxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Free Tier:

  • 2,000 free queries per month
  • Usually sufficient for research purposes

Configuration Steps

Set Environment Variables

export GITHUB_TOKEN=ghp_your_actual_github_token_here
export BRAVE_API_KEY=BSA_your_actual_brave_api_key_here

Verify Configuration

echo $GITHUB_TOKEN    # Should display your token
echo $BRAVE_API_KEY   # Should display your API key

FAQ

Q1: Why don't AI model keys need to be configured?

A: OpenCode and Claude Code already manage AI model API keys at the platform level. Once configured in the platform, all Agents can use them directly without needing to reconfigure for each project. Kilo Code (VS Code Plugin) typically uses your IDE's existing configuration.

Q2: What if I want to use a different search engine?

A: You can replace Brave Search MCP with other search MCPs, such as:

  • Google Search MCP (requires Google API Key)
  • Bing Search MCP (requires Bing API Key)
  • DuckDuckGo MCP (usually doesn't require API Key)

Q3: What permissions should I set for GitHub Token?

A:

  • Minimum: public_repo (access public repositories only)
  • Recommended: public_repo is sufficient for most research scenarios
  • Extended: Only need repo if analyzing private repositories

Q4: Is the Brave Search free tier sufficient?

A:

  • Free tier: 2,000 queries/month
  • One complete research: ~50-100 searches
  • Conclusion: Usually sufficient for personal research

Q5: What if my keys are leaked?

A:

  • GitHub Token: Immediately delete the old token at https://github.com/settings/tokens and generate a new one
  • Brave API Key: Regenerate in the Brave Dashboard
  • Check: Ensure keys are not committed to Git and are stored in a secret manager when possible

Security Best Practices

  1. Don't Commit to Git

    • Don't hardcode keys in code

  2. Regular Rotation

    • Recommend changing keys every 3-6 months
    • Especially GitHub Token

  3. Principle of Least Privilege

    • Only grant necessary permissions
    • Prefer public_repo for GitHub Token

  4. Monitor Usage

    • Regularly check GitHub Token usage
    • Monitor Brave Search API call counts

Troubleshooting

Issue: MCP Tools Not Working

Checklist:

# 1. Confirm environment variables are loaded
echo $GITHUB_TOKEN
echo $BRAVE_API_KEY

# 2. Restart the process that runs your platform/CLI if needed
# 3. Check key format
# GitHub Token should start with ghp_
# Brave API Key should start with BSA

Issue: GitHub MCP Returns 401 Unauthorized

Cause: Invalid GitHub Token or insufficient permissions

Solution:

  1. Check if token was copied correctly
  2. Confirm token hasn't expired
  3. Regenerate token

Issue: Brave Search Quota Exceeded

Cause: Exceeded free tier (2,000/month)

Solution:

  1. Wait for next month's reset
  2. Upgrade to paid plan
  3. Use alternative search MCP

Related Documentation

About

用于在不同 IDE \ CLI 工具之间进行 Agent 迁移的工具

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages