MCP Server Configuration

Some HVE Core agents use Model Context Protocol (MCP) servers to integrate with external services. MCP configuration is optional; agents that depend on MCP tools indicate when the required server is unavailable. The current installer flow and generated workspace template cover five servers: context7, microsoft-docs, ado, github, and figma.

TIP

The hve-core-installer skill included with HVE Core can generate MCP configuration with curated servers and guide you through peer clone, submodule, git-ignored, and other installation methods. Ask an agent to use the skill for guided setup. The sections below cover manual MCP configuration.

Overview

MCP tools extend GitHub Copilot's capabilities by connecting to external services. HVE Core references five curated MCP servers. Configure only the servers relevant to your workflow.

Choosing GitHub vs Azure DevOps

Most teams use one primary platform for repository hosting and work item management:

Repository Hosted On	Configure	Do Not Configure
GitHub	github server	ado server
Azure DevOps	ado server	github server
GitLab, Bitbucket, etc.	Neither	Both

Configuring both is unnecessary unless you work across platforms. If your repository is hosted elsewhere, configuration differs and is not documented here.

Backlog and work management does not require you to choose a platform up front. The backlog-plan and backlog-execute commands resolve the backing tracker at runtime from your workspace, so configure the server that matches the tracker you actually use. Jira and GitLab are reached through their own skills using credentials rather than an MCP server.

Agent MCP Dependencies

Agent, Prompt, or Skill	MCP Servers Used	Notes
backlog-plan	ado or github	Read-only backlog planning; server depends on the resolved tracker
backlog-execute	ado or github	Backlog mutations; server depends on the resolved tracker
Functional Planner	ado, microsoft-docs	PRD-to-work-item hierarchy planning
rpi-research	context7, microsoft-docs	Documentation lookup when available
RPI Agent	Varies by activated skill	Coordinates the applicable RPI phase skills
dt-figma-export	figma	DT artifact export to FigJam

Agents without MCP dependencies work without any MCP configuration.

Curated MCP Servers

HVE Core documents these five MCP servers:

context7

Library and SDK documentation lookup.

Property	Value
Type	stdio
Package	@upstash/context7-mcp

microsoft-docs

Microsoft Learn documentation access.

Property	Value
Type	http
URL	https://learn.microsoft.com/api/mcp

ado (Azure DevOps)

Azure DevOps work items, pipelines, and repositories.

Property	Value
Type	stdio
Package	@azure-devops/mcp
Requires	Organization name, optional tenant ID

github

GitHub repository and issue management.

Property	Value
Type	http
URL	https://api.githubcopilot.com/mcp/

figma

Figma and FigJam board creation and reading.

Property	Value
Type	http
URL	https://mcp.figma.com/mcp
Auth	Browser-based OAuth on first call
Local install	None
Used by	dt-figma-export prompt

The Figma MCP server requires a Figma account with a Dev or Full seat on a Professional, Organization, or Enterprise plan for sustained usage. Starter plans are limited to 6 tool calls per month. Authentication happens automatically via browser OAuth on first use.

Complete Configuration Template

Copy this template to .vscode/mcp.json in your workspace root. Remove servers you do not need (typically keep either github or ado, not both).

{
  "inputs": [
    {
      "id": "ado_org",
      "type": "promptString",
      "description": "Azure DevOps organization name (e.g. 'contoso')",
      "default": ""
    },
    {
      "id": "ado_tenant",
      "type": "promptString",
      "description": "Azure tenant ID (required for multi-tenant scenarios)",
      "default": ""
    }
  ],
  "servers": {
    "context7": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@upstash/context7-mcp"]
    },
    "microsoft-docs": {
      "type": "http",
      "url": "https://learn.microsoft.com/api/mcp"
    },
    "ado": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@azure-devops/mcp", "${input:ado_org}", "--tenant", "${input:ado_tenant}", "-d", "core", "work", "work-items", "search", "repositories", "pipelines"]
    },
    "github": {
      "type": "http",
      "url": "https://api.githubcopilot.com/mcp/"
    },
    "figma": {
      "type": "http",
      "url": "https://mcp.figma.com/mcp"
    }
  }
}

Method-Specific Notes

Extension Users

The VS Code extension provides agents without MCP configuration. Create .vscode/mcp.json in your project directory if you want to use MCP-dependent features.

Peer Clone / Submodule / Git-Ignored Users

Create .vscode/mcp.json in your workspace root (not inside the hve-core folder). VS Code reads MCP configuration only from the workspace root.

Codespaces / Devcontainer Users

Create .vscode/mcp.json in your repository's .vscode/ folder. The file will be available inside the container at the workspace root.

Multi-Root Workspace Users

MCP configuration can be placed in the .code-workspace file under settings or in the .vscode/mcp.json of the primary workspace folder. Workspace-level settings in the .code-workspace file take precedence.

Troubleshooting

Agent Reports MCP Tool Unavailable

1. Verify .vscode/mcp.json exists in workspace root
2. Check MCP server is running: View → Extensions → MCP SERVERS section
3. Trust the server when prompted by VS Code

Authentication Errors

• GitHub: Uses VS Code's built-in GitHub authentication
• ADO: Verify organization name and tenant ID are correct
• Figma: OAuth launches in your browser on first tool call. Re-authenticate by restarting VS Code if the session expires

MCP Server Not Starting

1. Ensure Node.js is installed and npx is available
2. Check the Output panel (View → Output → MCP Servers) for error messages
3. Verify network access to external URLs

References

• VS Code MCP Documentation
• GitHub MCP Server
• Azure DevOps MCP Server
• Microsoft Learn MCP Server
• Figma MCP Server

---

🤖 Crafted with precision by ✨Copilot following brilliant human instruction, then carefully refined by our team of discerning human reviewers.