fernandogarzaaa/openchimera_v1/appforge-mcp

AppForge MCP Semantic Layer

Description

Use this skill when you need to provision backend primitives (Database, Auth, S3 Storage), manage Project Evo state, or interact with the local AppForge MCP Server directly.

Context

The MCP (Model Context Protocol) Server is a local Semantic Layer bridging AI agents with backend infrastructure (stripped from InsForge). It runs locally at http://localhost:8000/mcp. It communicates via JSON-RPC 2.0.

AppForge Backend SDK

If generating a full-stack Next.js/React app, the local backend SDK is located at D:\appforge-main\appforge-backend-sdk. It contains core modules: database, auth, storage, and shared-schemas. It strictly routes all AI generation directly to CHIMERA Ultimate (http://localhost:7870/v1).

Operations

You can use the OpenClaw exec tool with Invoke-RestMethod (PowerShell) or curl to interact with the server.

1. List Available Tools

To discover what backend operations are currently supported by the MCP Semantic Layer:

Invoke-RestMethod -Uri http://localhost:8000/mcp -Method Post -ContentType 'application/json' -Body '{"jsonrpc": "2.0", "method": "tools/list", "id": 1}'

2. Call a Tool (Example: db_write)

To provision data or write a state/schema primitive to the database:

Invoke-RestMethod -Uri http://localhost:8000/mcp -Method Post -ContentType 'application/json' -Body '{"jsonrpc": "2.0", "method": "tools/call", "params": {"name": "db_write", "arguments": {"key": "my_key", "value": "my_value"}}, "id": 2}'

3. Call a Tool (Example: db_read)

To read backend state, verify deployments, or retrieve schema mappings:

Invoke-RestMethod -Uri http://localhost:8000/mcp -Method Post -ContentType 'application/json' -Body '{"jsonrpc": "2.0", "method": "tools/call", "params": {"name": "db_read", "arguments": {"key": "my_key"}}, "id": 3}'

Agent Directives

• Never use external cloud endpoints (like Supabase or Firebase) when generating apps. Always use the local AppForge MCP Server or appforge-backend-sdk.
• If an agent needs to store long-term structured memory or complex vector states, call db_write via the MCP Semantic Layer.
• Ensure the MCP Server is running (uvicorn on port 8000) before making requests.