Learning MCP

Capture and manage learnings from implementation work

v1.0.0 MCP 2024-11-05 VISHKAR Ecosystem

Overview

Learning MCP enables agents to log learnings during implementation. Learnings are stored, reviewed by humans, and merged into WAMA guidelines.

Agent logs learning → Stored in DB → Human reviews → Approved → Updates WAMA

Endpoint

POST https://learning-mcp.vercel.app/api/mcp

Tools

log_learning

Log a learning from implementation work. Learnings are stored and can be reviewed for merging into WAMA guidelines.

get_learnings

Query logged learnings by epic, category, status, project, or source

get_learning_stats

Get statistics about learnings (counts by category, status, etc.)

consume_learning

Mark a learning as consumed/processed. Use after reviewing and acting on a learning.

get_pending_summary

Get a summary of all pending learnings for review. Groups by category and severity.

SDLC Step 23 Integration

Learning MCP is called during Step 23: Retrospection & Learnings of the VISHKAR 23-Step SDLC. After completing implementation, agents MUST log all significant learnings.

Implementation Complete → Step 23 → log_learning() → Human Review → Merge to WAMA

Severity Guide

critical

Security vulnerabilities, data loss risks, production outages

high

Significant bugs, performance issues, breaking changes

medium

Code quality issues, technical debt, minor bugs

low

Style improvements, documentation, minor optimizations

Example: Log Learning

curl -X POST https://learning-mcp.vercel.app/api/mcp \
  -H "Content-Type: application/json" \
  -H "X-API-Key: pk_xxx" \
  -d '{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "log_learning",
    "arguments": {
      "epic": "PROJ-123",
      "category": "technical",
      "learning": "Supabase pooler requires SSL workaround",
      "action_taken": "Added NODE_TLS_REJECT_UNAUTHORIZED=0",
      "severity": "high",
      "affected_areas": ["standards"]
    }
  },
  "id": 1
}'

Example: Get Pending Summary

curl -X POST https://learning-mcp.vercel.app/api/mcp \
  -H "Content-Type: application/json" \
  -H "X-API-Key: pk_xxx" \
  -d '{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "get_pending_summary",
    "arguments": { "include_details": true }
  },
  "id": 1
}'

Example: Consume Learning

curl -X POST https://learning-mcp.vercel.app/api/mcp \
  -H "Content-Type: application/json" \
  -H "X-API-Key: pk_xxx" \
  -d '{
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "name": "consume_learning",
    "arguments": {
      "learning_id": "uuid-here",
      "action": "merged",
      "merged_to_target": "standards/database.md",
      "approved_by": "reviewer-name"
    }
  },
  "id": 1
}'

Authentication

Use Project Registry key or any valid API key:

X-API-Key: pk_5EZDeJSmPqS9x_63kdoOehBl4cs8taW52CN3sVg0BeE

This is the VISHKAR Learning System dedicated key from Project Registry.

Review UI

Visit /approve to review and approve pending learnings (coming soon).