- Adaptive Intelligence Framework integration - 94% TypeScript error reduction (103 → 6 errors) - 47 new runtime modules - 11 high-performance packages - 4,565 new tests added - Production readiness: 78/100 Resolving conflicts by accepting develop changes.
7.3 KiB
PCL Integration with AI Platforms
Overview
This guide shows how to integrate PCL with various AI platforms and assistants.
✅ MCP-Compatible Platforms (Native Support)
These platforms support the Model Context Protocol (MCP) natively and can use PCL directly without modification.
1. Claude Desktop ✅
Status: Full native MCP support
Setup:
-
Install PCL SDK:
npm install @pcl/sdk -
Create MCP Server: (
pcl-mcp-server.js)import { PclServer, StdioTransport, createRuntime, compile } from '@pcl/sdk'; import { readFileSync } from 'fs'; const source = readFileSync('./personas.pcl', 'utf-8'); const compiled = compile(source); const runtime = createRuntime(); runtime.load(compiled.value.program); const server = new PclServer({ name: 'pcl-server', version: '1.0.0', runtime, }); const transport = new StdioTransport(); server.connect(transport); -
Configure Claude Desktop:
Add to
~/Library/Application Support/Claude/claude_desktop_config.json(Mac) or%APPDATA%\Claude\claude_desktop_config.json(Windows):{ "mcpServers": { "pcl": { "command": "node", "args": ["/absolute/path/to/pcl-mcp-server.js"] } } } -
Restart Claude Desktop
-
Use PCL Personas:
You: List available PCL personas Claude: [Calls persona/list tool automatically] You: Execute the Analyst persona to analyze this data Claude: [Calls persona/execute with Analyst]
2. Claude Code (VS Code Extension) ✅
Status: Full MCP support
Setup:
-
Install Claude Code extension in VS Code
-
Create
.claude/mcp.jsonin your workspace:{ "mcpServers": { "pcl": { "command": "node", "args": ["./pcl-mcp-server.js"], "description": "PCL Persona Control Language" } } } -
Claude Code will automatically discover and use PCL tools
3. Cursor ✅
Status: MCP support available
Setup:
-
Create
cursor-mcp.json:{ "servers": { "pcl": { "command": "node", "args": ["./pcl-mcp-server.js"] } } } -
Configure in Cursor settings
-
Use PCL personas in Cursor chat
4. Zed Editor ✅
Status: MCP support
Setup:
Similar to Claude Code - configure MCP server in Zed settings.
5. Continue.dev ✅
Status: MCP compatible
Setup:
Configure in Continue.dev extension settings.
❌ Non-MCP Platforms (Requires REST API Wrapper)
These platforms do not support MCP natively. You need to create a REST API wrapper.
1. ChatGPT (OpenAI) ❌
Status: No MCP support - use REST API + Actions
Setup:
Step 1: Create REST API Server
Use the provided REST API wrapper:
# In examples/
node rest-api-wrapper.ts
This starts an HTTP server at http://localhost:3000 with endpoints:
GET /api/personas- List personasPOST /api/personas/:id/execute- Execute personaGET /api/teams- List teamsPOST /api/teams/:id/execute- Execute team
Step 2: Deploy REST API
Deploy to a public URL (Render, Railway, Fly.io, etc.):
# Example: Deploy to Render
git push render main
Get your public URL: https://your-pcl-api.onrender.com
Step 3: Create ChatGPT Action
- Go to ChatGPT → Create a GPT
- Click "Configure" → "Actions" → "Create new action"
- Import OpenAPI spec from
examples/openapi-spec.yaml - Update server URL to your deployed API
- Save and test
Step 4: Use in ChatGPT
You: List available PCL personas
ChatGPT: [Calls GET /api/personas]
You: Execute the Analyst persona with this data: [...]
ChatGPT: [Calls POST /api/personas/Analyst/execute]
Limitations:
- Requires separate REST API deployment
- Not as seamless as MCP
- Additional latency from HTTP calls
- Need to manage API authentication
2. DeepSeek ❌
Status: No MCP support - use REST API
Setup:
Same as ChatGPT above. DeepSeek would need to:
- Access your deployed REST API
- Use standard HTTP requests
- Parse JSON responses
Current Status: DeepSeek doesn't have a plugin/action system like ChatGPT, so integration is more limited.
3. Google Gemini ❌
Status: No MCP support - use REST API + Extensions
Setup:
Similar to ChatGPT - deploy REST API and configure as a Gemini Extension.
4. Perplexity AI ❌
Status: No MCP support
Limited integration options currently.
Comparison Matrix
| Platform | MCP Support | Integration Method | Difficulty | Latency |
|---|---|---|---|---|
| Claude Desktop | ✅ Native | MCP (stdio) | Easy | Low |
| Claude Code | ✅ Native | MCP (stdio) | Easy | Low |
| Cursor | ✅ Native | MCP (stdio) | Easy | Low |
| Zed | ✅ Native | MCP (stdio) | Easy | Low |
| Continue.dev | ✅ Native | MCP (stdio) | Easy | Low |
| ChatGPT | ❌ None | REST API + Actions | Medium | Medium |
| DeepSeek | ❌ None | REST API (manual) | Hard | Medium |
| Gemini | ❌ None | REST API + Extensions | Medium | Medium |
| Perplexity | ❌ None | Not available | N/A | N/A |
Recommended Approach by Use Case
For Development (Best Experience)
Use: Claude Desktop or Claude Code
- ✅ Native MCP support
- ✅ Lowest latency
- ✅ Best integration
- ✅ No deployment needed
For Production/Public Access
Use: REST API + ChatGPT Actions
- ✅ Wider accessibility
- ✅ Can serve multiple clients
- ⚠️ Requires deployment
- ⚠️ Additional latency
For Team Collaboration
Use: Cursor + MCP
- ✅ IDE integration
- ✅ Team can share MCP config
- ✅ Version control friendly
Future MCP Support
Platforms that might add MCP support:
- 🔮 ChatGPT (if OpenAI adopts MCP)
- 🔮 DeepSeek (if they implement MCP client)
- 🔮 GitHub Copilot (possible future integration)
- 🔮 Gemini (Google may support MCP)
MCP is an open protocol created by Anthropic, so adoption depends on each platform choosing to implement it.
Quick Decision Guide
Do you have Claude Desktop?
- YES → Use MCP (easiest!)
- NO → Continue below
Do you use VS Code?
- YES → Use Claude Code extension with MCP
- NO → Continue below
Do you need ChatGPT specifically?
- YES → Deploy REST API + create ChatGPT Action
- NO → Try Cursor or Zed with MCP
Do you want the simplest setup?
- Download Claude Desktop + configure MCP (5 minutes)
Summary
Best Option: Claude Desktop or Claude Code (native MCP support)
For ChatGPT/DeepSeek: You must create a REST API wrapper and deploy it publicly
Key Takeaway: MCP provides the best PCL integration experience, but you can still use PCL with non-MCP platforms through HTTP APIs.