Skip to main content

ADR-008: MCP Protocol Early Adoption

Status: DEFER to Phase 3
Date: 2025-08-25
Author: Architecture Review Team

Context​

The Model Context Protocol (MCP) is an emerging standard for AI agents to access external services. The plan includes building MCP server support from early stages to enable AI agents to query SmartBadge's semantic intelligence.

However, MCP is still new with limited adoption, and building for it early may be premature optimization.

Decision​

DEFER MCP implementation to Phase 3 (Month 3+) and focus on REST API first

Rationale:

  • MCP adoption is still limited (mainly Anthropic Claude)
  • REST API serves immediate customer needs
  • Can add MCP as additional interface later
  • Reduces MVP complexity significantly

Consequences​

Positive:

  • Simpler MVP architecture
  • Focus on proven API patterns
  • Faster time to market
  • Can validate demand before building MCP

Negative:

  • May miss early MCP adopter opportunity
  • Need to retrofit MCP support later
  • Potential competitive disadvantage if MCP takes off quickly
  • Loss of "AI-native" positioning

Alternatives Considered​

  1. MCP from Day One (Original Plan)

    • Pros: First-mover advantage, AI-native positioning
    • Cons: Complexity, limited initial users, unproven protocol

  2. REST API First, MCP Later (RECOMMENDED)

    • Pros: Proven approach, serves all clients, simpler
    • Cons: May miss MCP early adoption wave

  3. MCP Only

    • Pros: Fully AI-native, differentiated
    • Cons: Severely limited market, high risk

  4. Dual Implementation from Start

    • Pros: Covers all bases
    • Cons: Significant complexity, maintenance burden

Risk Assessment​

Risks of Deferring MCP:

  • Competitors could claim "AI-native" position
  • Miss opportunity to shape MCP ecosystem
  • Retrofit complexity if MCP patterns diverge from REST

Risks of Early MCP Adoption:

  • Significant development time for uncertain return
  • Protocol could change substantially
  • Limited client adoption initially

Migration Strategy​

Phase 1-2 (Months 1-2): REST API

// Standard REST API
app.get('/api/v1/repositories/:owner/:repo/analysis', async (req, res) => {
  const analysis = await getAnalysis(req.params)
  res.json(analysis)
})

app.post('/api/v1/search/similar', async (req, res) => {
  const results = await searchSimilar(req.body)
  res.json(results)
})

Phase 3 (Month 3+): Add MCP Layer

// MCP Server wrapping existing services
class SmartBadgeMCPServer {
  async handleRequest(request: MCPRequest): Promise<MCPResponse> {
    // Wrap existing REST endpoints
    switch (request.method) {
      case 'search':
        return this.wrapRestEndpoint('/api/v1/search/similar', request)
      case 'analyze':
        return this.wrapRestEndpoint('/api/v1/repositories/analyze', request)
    }
  }
}

Success Criteria for MCP Adoption​

Before implementing MCP, validate:

  1. Demand Signal: >10 customers requesting MCP access
  2. Protocol Stability: MCP v1.0 released
  3. Client Adoption: >3 major AI platforms supporting MCP
  4. Revenue Opportunity: Clear monetization path for MCP access

Conclusion​

While MCP represents an exciting future for AI-service integration, the protocol's immaturity and limited adoption make it a poor fit for MVP. Focus on delivering value through proven REST APIs first, then add MCP support when market demand is validated. This approach reduces risk while maintaining optionality for future AI-native features.