AI Agent Integration

Juicebox provides comprehensive tooling for AI assistants and automated development workflows. This includes an MCP server for documentation access, Claude Code skills for protocol operations, and machine-readable data files.

Claude Code Skills

The fastest way to build on Juicebox with AI assistance is using our Claude Code skills plugin.

Install: Add to your Claude Code settings or clone github.com/mejango/juicebox-skills

Available Skills

Skill	Description
/jb-project	Create and configure Juicebox V5 projects
/jb-ruleset	Configure and queue rulesets with all parameters
/jb-pay-hook	Generate custom pay hooks with Foundry tests
/jb-cash-out-hook	Generate custom cash out hooks with tests
/jb-split-hook	Generate custom split hooks with tests
/jb-decode	Decode and analyze transaction calldata
/jb-query	Query project state from the blockchain
/jb-docs	Search documentation via MCP server
/jb-v5-api	API reference for all contract functions
/jb-v5-impl	Deep implementation knowledge and edge cases

Example Usage

/jb-project

Create a crowdfunding project that:
- Accepts ETH
- Issues 1M tokens per ETH
- Reserves 20% for the team
- Has a 100 ETH payout limit per 30-day cycle
- Allows token holders to cash out at any time

The skill will generate complete deployment code with proper configurations.

---

Machine-Readable Resources

For programmatic access to protocol data:

Resource	URL	Description
LLM Summary	/llms.txt	Quick protocol overview for LLMs
Full Context	/llms-full.txt	Complete protocol reference
Contract Addresses	/api/contracts.json	All V5 addresses as JSON
SDK Reference	/api/sdk.json	juice-sdk-react hooks and types

---

MCP Server

The Juicebox documentation is available through an MCP (Model Context Protocol) server for AI assistants.

HTTP API

Base URL: https://docs.juicebox.money/api/mcp

Documentation Endpoints

Method	Endpoint	Description
POST	/search	Search docs by query, category, version
POST	/get-doc	Get full document by path or title
GET	/list-docs	List all docs in a category
GET	/structure	Get documentation structure/stats

Code & Integration Endpoints

Method	Endpoint	Description
POST	/search-code	Search code examples by language
GET	/contracts	Get contract addresses by name/chain
GET	/sdk	Get SDK hook/utility reference
GET	/patterns	Get integration patterns by project type

Quick Examples

Search documentation:

curl -X POST https://docs.juicebox.money/api/mcp/search \
  -H "Content-Type: application/json" \
  -d '{"query": "deploy project", "category": "developer", "limit": 5}'

Get contract addresses:

curl "https://docs.juicebox.money/api/mcp/contracts?contract=JBController"

Get integration pattern:

curl "https://docs.juicebox.money/api/mcp/patterns?projectType=revnet"

Search code examples:

curl -X POST https://docs.juicebox.money/api/mcp/search-code \
  -H "Content-Type: application/json" \
  -d '{"query": "pay", "language": "solidity", "limit": 5}'

Parameters

Documentation Search:

• query (string, required): Search terms
• category: developer | user | dao | ecosystem | all
• version: v3 | v4 | v5 | all
• limit (number): Max results (default: 10)

Contract Addresses:

• contract: Contract name (e.g., JBController, REVDeployer)
• chainId: 1 | 10 | 8453 | 42161 | testnets | all
• category: core | revnet | hooks | suckers | omnichain | all

SDK Reference:

• package: juice-sdk-react | juice-sdk-core | revnet-sdk | all
• hook: Specific hook name (e.g., useJBProjectProvider)
• category: context | read | write | omnichain | math | all

Integration Patterns:

• pattern: Pattern ID (e.g., pay-project, wagmi-setup)
• projectType: crowdfunding | revnet | dao-treasury | subscription

---

Using with Claude Desktop

For local MCP server access:

1. Clone and install:

   git clone https://github.com/jbx-protocol/juice-docs-v3
   cd juice-docs-v3
   npm run mcp:install && npm run mcp:build-index

2. Configure Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json):

   {
     "mcpServers": {
       "juice-docs": {
         "command": "node",
         "args": ["/path/to/juice-docs-v3/mcp-server/src/index.js"],
         "cwd": "/path/to/juice-docs-v3"
       }
     }
   }

3. Restart Claude Desktop

---

For AI Agent Developers

If you're building AI agents that interact with Juicebox:

Recommended Integration Order

1. Use Claude Code Skills - Most comprehensive, handles complex operations
2. Query MCP Server - For documentation lookup and code examples
3. Use JSON APIs - For contract addresses and SDK reference
4. Parse llms-full.txt - For complete protocol context

Key Capabilities

Task	Recommended Approach
Deploy a project	/jb-project skill
Generate custom hook	/jb-pay-hook, /jb-cash-out-hook, or /jb-split-hook
Query project state	/jb-query skill with cast/ethers
Decode transaction	/jb-decode skill
Look up function signature	/jb-v5-api skill or MCP /get-doc
Understand internal mechanics	/jb-v5-impl skill
Get contract address	MCP /contracts or /api/contracts.json

Example: AI Agent Workflow

// 1. Get contract addresses
const contracts = await fetch('https://docs.juicebox.money/api/contracts.json')
  .then(r => r.json());

// 2. Search for relevant documentation
const docs = await fetch('https://docs.juicebox.money/api/mcp/search', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ query: 'launch project', category: 'developer' })
}).then(r => r.json());

// 3. Get specific document
const guide = await fetch('https://docs.juicebox.money/api/mcp/get-doc', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ path: 'dev/v5/build/life-of-a-project.md' })
}).then(r => r.json());

---

More Information

• Claude Code Skills Repository
• MCP Server README
• MCP Specification
• LLM Quick Reference
• Full Protocol Context