ms-ai-architect/skills/ms-ai-advisor/references/architecture/recommended-mcp-servers.md

# Recommended MCP Servers for AI Architect

**Last updated:** 2026-02
**Status:** Advisory
**Category:** Architecture

---

## Introduksjon

MCP (Model Context Protocol) servers extend the AI Architect plugin by providing real-time access to external tools and data sources. Rather than relying solely on static knowledge base files, MCP servers let the architect agent query live documentation, manage infrastructure, generate diagrams, and interact with project management systems directly during a session.

This reference documents which MCP servers are already integrated, which are recommended for enhanced functionality, and how they map to the architect workflow phases.

---

## Allerede integrert

| Server | Formål | Tools | Workflow Phase |
|--------|--------|-------|----------------|
| `microsoft-learn` | Offisiell Microsoft dokumentasjon | `microsoft_docs_search`, `microsoft_docs_fetch`, `microsoft_code_sample_search` | Knowledge Validation (Phase 4-5) |
| `mcp-image` | Bildegenerering med Imagen 3 | `generate_image` | Visualization (Phase 7) |

### microsoft-learn
Primary knowledge validation tool. Used by the research-agent to fetch latest platform capabilities, pricing, regional availability, and best practices. Critical for ensuring recommendations are current.

**Typical usage in architect workflow:**
- Verify service availability in Norway East/West regions
- Check latest SDK versions and deprecation notices
- Validate security configuration recommendations
- Fetch code samples for POC plans

### mcp-image
Used by the diagram-generation-agent to create architecture diagrams via Imagen 3. Produces visual representations of proposed architectures for documentation and stakeholder communication.

---

## Anbefalte tillegg

### Azure MCP Server (microsoft/azure-mcp-server)

**Description:** Official Microsoft MCP server for Azure resource management. Provides read/write access to Azure subscriptions, resource groups, and individual services.

**Key tools:**
- Resource group listing and management
- Service configuration inspection
- Deployment status checking
- Cost and usage data retrieval

**Use cases for architect:**
- Validate existing infrastructure before proposing changes
- Check current SKUs and configuration for cost optimization reviews
- Verify network topology and security group rules during security assessments
- Inspect AI service deployments (Azure OpenAI endpoints, AI Search indexes)
- Compare proposed architecture against actual deployed state

**Relevant commands:** `/architect:security`, `/architect:cost`, `/architect:review`

**Installation:**
```json
{
  "mcpServers": {
    "azure": {
      "command": "npx",
      "args": ["-y", "@azure/mcp-server"],
      "env": {
        "AZURE_SUBSCRIPTION_ID": "<your-subscription-id>"
      }
    }
  }
}
```

---

### Bicep MCP Server

**Description:** Infrastructure as Code generation and validation for Azure using Bicep templates. Translates architecture decisions into deployable infrastructure definitions.

**Key tools:**
- Bicep template generation from natural language
- Template validation and what-if analysis
- Parameter file generation
- Module composition

**Use cases for architect:**
- Generate IaC templates from ADR decisions (`/architect:adr` output)
- Validate proposed infrastructure is deployable
- Create POC infrastructure templates (`/architect:poc` output)
- Ensure compliance with Azure Policy through template validation
- Generate migration scripts for `/architect:migrate` plans

**Relevant commands:** `/architect:adr`, `/architect:poc`, `/architect:migrate`

**Installation:**
```json
{
  "mcpServers": {
    "bicep": {
      "command": "npx",
      "args": ["-y", "@azure/bicep-mcp-server"]
    }
  }
}
```

---

### Azure DevOps MCP Server (microsoft/azure-devops-mcp)

**Description:** Integration with Azure DevOps for work items, pipelines, repositories, and boards. Bridges architecture decisions with implementation tracking.

**Key tools:**
- Work item creation and querying
- Pipeline status and trigger
- Repository browsing
- Board and sprint management

**Use cases for architect:**
- Create implementation work items from architecture review findings
- Track ADR implementation progress
- Link POC plans to sprint backlogs
- Monitor deployment pipeline status for migration plans
- Query existing codebase for integration point analysis

**Relevant commands:** `/architect:review`, `/architect:poc`, `/architect:migrate`

**Installation:**
```json
{
  "mcpServers": {
    "azure-devops": {
      "command": "npx",
      "args": ["-y", "@microsoft/azure-devops-mcp"],
      "env": {
        "AZURE_DEVOPS_ORG": "<your-org>",
        "AZURE_DEVOPS_PAT": "<your-pat>"
      }
    }
  }
}
```

---

### Playwright MCP Server

**Description:** Browser automation for visual testing and verification. Enables the architect plugin to visually verify deployed solutions and capture screenshots.

**Key tools:**
- Page navigation and screenshot capture
- Element interaction and form filling
- Visual regression comparison
- Network request interception

**Use cases for architect:**
- Visual verification of diagram-generation-agent output
- Screenshot capture of Azure Portal configurations during reviews
- Validate Copilot Studio agent behavior in browser
- Capture evidence for architecture review documentation
- Accessibility testing (WCAG compliance checks)

**Relevant commands:** `/architect:diagram`, `/architect:review`

**Installation:**
```json
{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["-y", "@anthropic/mcp-playwright"]
    }
  }
}
```

---

## MCP Server Selection Matrix

| Workflow Need | Primary MCP | Fallback |
|---------------|-------------|----------|
| Documentation lookup | microsoft-learn | WebSearch |
| Resource inspection | azure | Azure Portal (manual) |
| IaC generation | bicep | Manual Bicep authoring |
| Work item tracking | azure-devops | Linear (already configured) |
| Visual verification | playwright | Manual screenshot |
| Diagram generation | mcp-image | Mermaid in markdown |

---

## Installasjon

Add MCP servers to your Claude Code settings file at `~/.claude/settings.json` or project-level `.claude/settings.json`:

```json
{
  "mcpServers": {
    "microsoft-learn": {
      "command": "npx",
      "args": ["-y", "@anthropic/mcp-microsoft-learn"]
    },
    "azure": {
      "command": "npx",
      "args": ["-y", "@azure/mcp-server"],
      "env": {
        "AZURE_SUBSCRIPTION_ID": "<sub-id>"
      }
    },
    "bicep": {
      "command": "npx",
      "args": ["-y", "@azure/bicep-mcp-server"]
    }
  }
}
```

**Notes:**
- MCP servers requiring authentication (Azure, Azure DevOps) need environment variables configured
- Use `.env` files or secret managers -- never commit credentials
- Test each server independently before combining
- Monitor MCP server resource usage in long sessions

---

## For Cosmo

These MCP servers enhance the 7-phase architect workflow:

| Phase | MCP Enhancement |
|-------|-----------------|
| 1. Problem Understanding | azure-devops: Query existing work items and requirements |
| 2. Context & Constraints | azure: Inspect current infrastructure state |
| 3. Capacity & Ambition | azure: Check subscription limits and quotas |
| 4. Knowledge Validation | microsoft-learn: Verify latest documentation |
| 5. Knowledge Integration | microsoft-learn + azure: Combine docs with live state |
| 6. Architecture Proposal | bicep: Generate deployable IaC from proposal |
| 7. Visualization | mcp-image: Generate architecture diagrams |

**Priority order for adoption:**
1. `microsoft-learn` (already integrated, essential)
2. `mcp-image` (already integrated, visualization)
3. `azure` (highest value-add for live infrastructure validation)
4. `bicep` (IaC generation from architecture decisions)
5. `azure-devops` (implementation tracking bridge)
6. `playwright` (visual verification, nice-to-have)