Using IconVectors with GitHub Copilot CLI

Overview

GitHub Copilot CLI can use IconVectors 2.00 through a local stdio MCP server on Windows. Copilot CLI launches IconVectorsMcp.exe on the user’s machine; the sidecar forwards calls over loopback to the already-running desktop application.

This is separate from Copilot Chat in Visual Studio Code, which uses .vscode/mcp.json and is documented in Using IconVectors with VS Code / GitHub Copilot. See MCP Integration for the shared IconVectors contract.

Prerequisites

  • Install GitHub Copilot CLI on the same Windows machine as IconVectors.

  • Start IconVectors 2.00 normally and leave it running.

  • Confirm IconVectorsMcp.exe exists beside IconVectors.exe.

  • The default bridge is 127.0.0.1:61337. Match a custom Options/McpPort value if necessary.

Configure Copilot CLI MCP

Add the user-scoped server with the recommended name iconvectors and expose all tools:

copilot mcp add --tools "*" iconvectors -- "C:\Program Files\Axialis\IconVectors\IconVectorsMcp.exe" --port 61337
copilot mcp list
copilot mcp get iconvectors

The path is the normal installed-product example. Replace it or the port when needed. Persistent user configuration is ~/.copilot/mcp-config.json (%USERPROFILE%\.copilot\mcp-config.json on Windows).

Equivalent JSON uses this shape:

{
  "mcpServers": {
    "iconvectors": {
      "type": "stdio",
      "command": "C:\\Program Files\\Axialis\\IconVectors\\IconVectorsMcp.exe",
      "args": ["--port", "61337"],
      "tools": ["*"]
    }
  }
}

A project can use .mcp.json for local checkout configuration or .github/mcp.json for a shared file. Project servers require folder trust. Copilot CLI does not directly read VS Code’s .vscode/mcp.json.

Verify the connection

Use copilot mcp list and copilot mcp get iconvectors. In an interactive session, use /mcp list, /mcp show iconvectors, and /mcp reload. Then call:

  • app_ping

  • app_getInfo

  • app_getCapabilities

  • app_get_workspace

Inspect the active document and selection or Explorer state before making a change.

Install the IconVectors skill and instructions

Copy the complete distributed iconvectors-mcp skill, including its references directory, to one of these project locations:

  • .agents/skills/iconvectors-mcp/SKILL.md

  • .github/skills/iconvectors-mcp/SKILL.md

  • .claude/skills/iconvectors-mcp/SKILL.md

Use /skills reload and /skills info iconvectors-mcp to verify it. Copilot CLI also reads AGENTS.md and GitHub instruction files. Keep those files concise and leave task-specific expertise and schemas in the skill.

First workflow

Try one inspected change:

  1. Inspect application capabilities, workspace, document, and selection.

  2. Make one bounded Editor mutation.

  3. Reinspect the affected element and document.

  4. Save or export only when a native Windows destination was requested.

Use a history transaction for related active-document edits. Document replacement, save/export, and Explorer filesystem operations are not covered by document Undo.

Working with 155 tools

GitHub Copilot CLI is naming-compatible with 155/155 production IconVectors tools. With the server name iconvectors, no sanitization, truncation, collision, or client-specific rename is required.

The server’s tools list accepts "*" or selected raw MCP tool names. Session options --available-tools and --excluded-tools can narrow exposure. Leave deferred tools at the default deferTools: "auto" so Copilot can load a large catalog on demand; "never" keeps all definitions visible. Use /context to inspect pressure and /compact when conversation history grows.

These are Copilot CLI controls, not IconVectors server profiles.

Tool approval

MCP invocations require permission by default. Narrow controls can allow a read-only health check while denying a replacement tool:

copilot --allow-tool="iconvectors(app_ping)" --deny-tool="iconvectors(document_setSvg)"

Deny rules override allow rules. Keep prompts for save/export, replacement, Explorer file operations, and batch apply. Do not use blanket server approval merely to remove friction.

Troubleshooting

Server is absent

Check copilot mcp list, copilot mcp get iconvectors, scope, trust, absolute command path, port, and the tools filter.

Discovery appears stale

Run /mcp reload. Use the documented cache-disable setting only if live discovery still appears stale.

Bridge is unavailable

Start IconVectors, match port 61337, and verify the intended application, workspace, and document.

Logs are needed

Inspect ~/.copilot/logs/; /session reports the current log path. Use --log-level=debug or /diagnose for bounded troubleshooting.

Path or output fails

Use native Windows paths and bounded DOM, SVG, Explorer, or preview results.

Copilot CLI versus the cloud coding agent

Copilot CLI runs locally and can launch the local sidecar. GitHub’s Copilot cloud coding agent runs in a separate hosted environment. Its localhost is not the user’s Windows desktop, so it cannot directly reach this IconVectors stdio/loopback integration. Cloud-agent MCP configuration is not an alternate route to the live desktop application.

For current CLI commands, see the official GitHub Copilot CLI MCP documentation.