internal/docs/ROADMAP_2026.md
Last Updated: March 2026
The emergence of AI agents represents a paradigm shift in how services are consumed. Where APIs served apps, MCP serves agents. Go Micro is uniquely positioned to become the standard microservices framework for the agent era.
This roadmap outlines Go Micro's evolution from an API-first framework to an AI-native platform while maintaining backward compatibility and ensuring long-term sustainability.
┌──────────┐ HTTP/REST ┌─────────────┐ RPC ┌──────────┐
│ Mobile │ ───────────────→ │ Gateway │ ─────────→ │ Services │
│ App │ │ (Express) │ │ │
└──────────┘ └─────────────┘ └──────────┘
Characteristics:
┌──────────┐ MCP/SSE ┌─────────────┐ RPC ┌──────────┐
│ Claude │ ───────────────→ │ MCP │ ─────────→ │ Services │
│ GPT │ │ Gateway │ │ │
└──────────┘ └─────────────┘ └──────────┘
Characteristics:
API Gateways solve integration for developers. MCP solves integration for AI.
Go Micro's MCP integration means:
Make every microservice AI-native by default.
micro run enables MCP automaticallyStatus: COMPLETE as of February 2026
gateway/mcp)micro run --mcp-address)Status: COMPLETE (100%) - All core features and documentation delivered
Theme: Make it trivial for any AI to call your services
micro mcp command for Claude Code integrationWhy: Claude Code and other local AI tools use stdio MCP servers. This enables:
# In Claude Code config
{
"mcpServers": {
"my-services": {
"command": "micro",
"args": ["mcp"]
}
}
}
Business value: Direct integration with Anthropic's flagship developer tool.
@param, @return, @exampleBefore:
Tools:
- users.Users.Get - Call Get on users service
After:
Tools:
- users.Users.Get
Description: Retrieve user profile by ID. Returns full profile including email,
name, created date, and preferences.
Parameters:
- id (string, required): User ID in UUID format
Returns: User object with profile fields
Example: {"id": "123e4567-e89b-12d3-a456-426614174000"}
Why: Better descriptions = better agent performance. Agents need context to call services correctly.
Why: Different agents prefer different protocols. Support them all.
Create official SDKs for popular agent frameworks:
langchain-go-micro Python packageai.Model interface with Generate and Streamai/anthropic)ai/openai)Why: The ai package powers the agent playground and enables services to call AI models directly.
go-micro-llamaindex packageBusiness value: Every agent framework can use go-micro services out of the box.
micro mcp Command Suite ✅ COMPLETEImplemented:
# Start MCP server
micro mcp serve # Stdio (for Claude Code) ✅
micro mcp serve --address :3000 # HTTP/SSE (for web agents) ✅
# Development
micro mcp list # List available tools ✅
micro mcp list --json # JSON output ✅
micro mcp test users.Users.Get # Test a tool ✅
micro mcp docs # Generate MCP documentation ✅
micro mcp docs --format json # JSON output ✅
micro mcp export langchain # Export to LangChain format ✅
micro mcp export openapi # Export as OpenAPI ✅
micro mcp export json # Export as JSON ✅
micro run dashboardExample:
http://localhost:8080/playground
> You: "Show me user 123's last 5 orders"
Agent: Let me check that...
→ Calling users.Users.Get with {"id": "123"}
→ Calling orders.Orders.List with {"user_id": "123", "limit": 5}
Here are the 5 most recent orders for Alice Smith:
1. Order #45678 - $125.00 - Shipped (Jan 15)
2. Order #45123 - $89.99 - Delivered (Jan 10)
...
Business value: Instant demos. Show investors/customers AI calling your services.
Status: IN PROGRESS (50%) - Core security and observability features delivered early, infrastructure work remaining
Theme: Run MCP gateways in production at scale
Create a production-grade standalone MCP gateway:
micro-mcp-gatewayDeployment:
# Standalone gateway
micro-mcp-gateway \
--registry consul:8500 \
--address :3000 \
--auth jwt \
--rate-limit 1000/hour \
--cache redis:6379
Business value: Enterprise customers need production-grade MCP gateways. This is a paid offering.
Dashboard Example:
Agent Activity - Last 7 Days
─────────────────────────────
Claude Desktop 1,234 calls $12.34 compute cost
ChatGPT Plugin 567 calls $5.67 compute cost
Custom Agent 234 calls $2.34 compute cost
Top Services
────────────
users 45%
orders 30%
payments 15%
Slowest Tools
─────────────
analytics.Reports.Generate 2.3s avg
payments.Payments.Process 890ms avg
Business value: Enterprises need observability. This justifies MCP Gateway pricing.
Implemented Example:
mcp.Serve(mcp.Options{
Registry: registry,
Auth: authProvider, // ✅ Implemented
Scopes: map[string][]string{ // ✅ Implemented
"blog.Blog.Create": {"blog:write"},
"blog.Blog.Delete": {"blog:admin"},
},
AuditFunc: func(r mcp.AuditRecord) { // ✅ Implemented
log.Printf("[audit] %+v", r)
},
})
Implemented - Metadata in Context:
// Trace ID, Tool Name, and Account ID are automatically
// propagated to services via context metadata:
// - Mcp-Trace-Id
// - Mcp-Tool-Name
// - Mcp-Account-Id
Future Enhancement - Service-Side Example:
// Future: Direct access to agent info from context
func (s *Users) Delete(ctx context.Context, req *Request, rsp *Response) error {
// For now, services can read metadata keys:
// Mcp-Account-Id, Mcp-Trace-Id, Mcp-Tool-Name
md, _ := metadata.FromContext(ctx)
accountID := md["Mcp-Account-Id"]
if accountID != "admin-account" {
return errors.Forbidden("users", "admin only")
}
// ...
}
Business value: Security is a hard requirement for enterprise adoption.
micro-operator for KubernetesMCPGateway resourceExample:
apiVersion: micro.dev/v1
kind: MCPGateway
metadata:
name: production-gateway
spec:
registry: consul
replicas: 3
rateLimit:
perAgent: 1000/hour
observability:
otel: true
traces: jaeger:14268
deploy/helm/mcp-gateway/)Business value: Easy deployment = faster adoption.
Target: Support 10,000 concurrent agent requests on a single gateway.
Theme: Build the MCP ecosystem and sustainable business
Create a marketplace of pre-built AI agents that use go-micro services:
Developers build agents that solve specific problems using microservices:
Examples:
Format:
# agent.yaml
name: customer-support
description: AI agent that handles customer support tickets
services:
- users
- tickets
- orders
- payments
prompts:
- system: "You are a helpful customer support agent..."
- examples: [...]
mcp:
gateway: "mcp://services.company.com"
pricing: free|paid
Usage:
# Install agent from marketplace
micro agent install customer-support
# Run agent
micro agent run customer-support
# Agent now has access to your services via MCP
Business value:
Build a sustainable business model around open-source core:
go-micro.dev/v5)gateway/mcp)micro run, micro server)Target: Teams that want managed MCP gateways
Features:
Pricing:
Value proposition: "Don't run your own MCP gateway. We'll do it for you."
Target: Large companies deploying at scale
Features:
Pricing:
Value proposition: "Production-grade MCP for your entire organization."
Pricing: $200-300/hour
Why: Anthropic created MCP. Being their preferred microservices framework drives adoption.
Why: OpenAI has largest AI user base. Tap into that market.
Business value: Every major AI platform can use go-micro services.
Target: Grow from 5K GitHub stars to 15K+ by end of 2026.
Theme: The AI-native microservices platform
Go Micro becomes the platform layer between AI and infrastructure:
┌─────────────────────────────────────┐
│ AI Agents Layer │
│ Claude | GPT | Gemini | Custom │
└─────────────────────────────────────┘
↓ MCP
┌─────────────────────────────────────┐
│ Go Micro Platform │
│ Gateway | Registry | Auth | Mesh │
└─────────────────────────────────────┘
↓ RPC
┌─────────────────────────────────────┐
│ Microservices Layer │
│ Users | Orders | Payments | ... │
└─────────────────────────────────────┘
Example:
$ micro generate "a user authentication service with JWT"
[AI] Analyzing requirements...
[AI] Generating service scaffold...
[AI] Adding JWT auth with RS256...
[AI] Creating database schema...
[AI] Writing tests...
[AI] Service ready: ./auth-service
$ cd auth-service && micro run
[AI] Service running. MCP-enabled. Try asking Claude to create a user!
Target ARR: $1M Year 1, $5M Year 2
Customer Segments:
Unit Economics:
Growth Strategy:
Target ARR: $500K Year 1, $3M Year 2
Value Proposition:
Typical Deal:
Target Revenue: $250K Year 1, $750K Year 2
Services:
Pricing:
Target Revenue: $100K Year 1, $500K Year 2
Model:
Growth:
Year 1 (2026): $1.85M
Year 2 (2027): $9.25M (5x growth)
Year 1 (Lean):
Year 2 (Growth):
Why such high margins?
Recommendation: Bootstrap first, then raise Series A if needed for expansion.
Risk: MCP protocol changes (Anthropic controls spec)
Risk: Performance issues at scale
Risk: Security vulnerabilities in MCP gateway
Risk: AI hype dies down
Risk: Competitors build MCP support
Risk: Cloud providers offer competing solutions
Risk: Enterprises slow to adopt agents
Risk: Different MCP implementations fragment market
Our advantage: First-mover in MCP + microservices space.
Our advantage: Purpose-built for agent integration, not retrofitted.
Defense:
The future of microservices is AI-native.
API gateways connected apps to services. MCP connects agents to services.
Go Micro is uniquely positioned to own this space:
The question isn't whether agents will use microservices. The question is: which framework will they use?
Let's make it Go Micro.
Next Steps (March 2026):
Questions? Feedback?
This roadmap is a living document. It will evolve based on market feedback, technical discoveries, and community input. Last updated: March 2026.