ai-assistant: Add ability to run mcp server from ai-assistant plugin

[ashu8912]

No description provided.

[illume]

It's important to do these standard things for PRs to speed reviews along:

• PR description
• screenshots or video
• steps to test each fix or piece of functionality
• link to the related issue
• If there are no tests at all, please justify this. Note that it will slow reviews down, and slow down future changes
• Document fields and functions. This speeds up the review, and helps speed up further changes.
• git commit messages with context of the modules changed in subject, AND a "why" in the git commit body unless it's covered in the git message subject.

ai-assistant: MCPSettings: Add MCP settings UI and Electron config API

- Add components/settings/MCPSettings.tsx: UI for configuring MCP servers (add/remove/toggle/edit),
  uses window.desktopApi.mcp.getConfig/updateConfig when running in Electron, falls back to pluginStore.
- Extend ElectronMCPApi typings with getConfig and updateConfig for config persistence.
- Provide default server when enabling MCP with no servers, and a "Test MCP Connection" action.

Notes:
- Argument parsing is space-based (no quote handling).
- Ensure native side exposes getTools (used by the test action) and that updateConfig triggers any required client restart.

[Copilot]

Pull Request Overview

This PR adds comprehensive MCP (Model Context Protocol) integration to the AI assistant, enabling connection to external tools and services. It introduces tool approval workflows, intelligent argument processing, and enhanced tool management capabilities.

• Added MCP server configuration and management in Electron environments
• Implemented tool approval dialogs with inline confirmation UI
• Created intelligent argument processing using AI for MCP tools
• Added tool orchestration for multi-tool execution

Reviewed Changes

Copilot reviewed 40 out of 42 changed files in this pull request and generated 13 comments.

Show a summary per file

File	Description
ai-assistant/src/utils/ToolConfigManager.ts	Added source tracking ('built-in' vs 'mcp'), MCP tool integration functions, and tool state initialization
ai-assistant/src/utils/ToolApprovalManager.ts	New approval manager for tool execution with session-based auto-approval
ai-assistant/src/utils/InlineToolApprovalManager.ts	Inline tool approval with user context extraction for intelligent argument processing
ai-assistant/src/langchain/tools/ToolManager.ts	Enhanced with MCP tool support, argument mapping, and formatter integration
ai-assistant/src/langchain/tools/ToolOrchestrator.ts	New orchestrator for analyzing and coordinating multi-tool execution
ai-assistant/src/langchain/formatters/MCPOutputFormatter.ts	AI-powered formatter for MCP tool output with markdown support
ai-assistant/src/langchain/LangChainManager.ts	Major updates for tool orchestration, inline approval, and argument enhancement
ai-assistant/src/components/mcpOutput/MCPOutputDisplay.tsx	Rich UI component for displaying formatted MCP tool results
ai-assistant/src/components/mcpOutput/MCPArgumentProcessor.ts	Intelligent argument processing with schema validation
ai-assistant/src/components/settings/MCPSettings.tsx	MCP server configuration UI with enable/disable controls
ai-assistant/src/components/common/InlineToolConfirmation.tsx	Inline tool confirmation component (exported via index)
ai-assistant/src/types/electron.d.ts	TypeScript definitions for Electron MCP API
ai-assistant/src/hooks/useClusterChangeNotifier.ts	Hook for monitoring cluster changes and notifying MCP servers

Files not reviewed (1)

• ai-assistant/package-lock.json: Language not supported

Comments suppressed due to low confidence (2)

ai-assistant/src/langchain/tools/ToolManager.ts:226

• This statement is unreachable.
  ai-assistant/src/langchain/tools/ToolManager.ts:258
• This statement is unreachable.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Duplicate return statement on lines 225-226. The second return is unreachable and should be removed.

[Copilot]

Duplicate return statement on lines 257-258. The second return is unreachable and should be removed.

[Copilot]

Unused import Paper.

[Copilot]

Unused import Chip.

[Copilot]

The initial value of type is unused, since it is always overwritten.

Suggested change

	let type: FormattedMCPOutput['type'] = 'text';
	let type: FormattedMCPOutput['type'];

[Copilot]

This use of variable 'isDenying' always evaluates to false.

[Copilot]

This use of variable 'isApproving' always evaluates to false.

[Copilot]

This use of variable 'isApproving' always evaluates to false.

[Copilot]

This use of variable 'isApproving' always evaluates to false.

[Copilot]

This use of variable 'args' always evaluates to true.

[illume]

I'm working on this.

[illume]

Wtih Azure OpenAI provider gpt4 it's having issues.
With this plugin and headlamp branch mcp-electron 0b95d6d9d38445456655080a042badc3a698689c

With ai-assistant installed from Plugin Catalog it's working.

https://mgns-eastus2.openai.azure.com/openai/deployments/gpt-4/chat/completions?api-version=2024-12…
Request Method: POST

Status Code: 400 Bad Request
{
   "error": {
       "message": "'Make requests to the Kubernetes API server to fetch, create, update or delete resources.\n\nRESOURCE UPDATE GUIDELINES:\n- For UPDATE/MODIFY/CHANGE operations: Use PUT method with ONLY the specific fields to change\n- Provide patch objects that will be merged with existing resources\n- Use null values to remove fields (e.g., {"spec": {"livenessProbe": null}})\n- The system automatically merges patches with current resources before making PUT requests\n\nLOG HANDLING FOR MULTI-CONTAINER PODS:\n- When a user asks for logs from a pod, ALWAYS first check the pod specification to determine the number of containers\n- If the pod has only one container, directly fetch logs: kubernetes_api_request(url="/api/v1/namespaces/default/pods/pod-name/log", method="GET")\n- If the pod has multiple containers, you MUST ask the user which container they want logs from\n- List the available container names and ask the user to specify\n- Once the user specifies a container, fetch logs with the container parameter: kubernetes_api_request(url="/api/v1/namespaces/default/pods/pod-name/log?container=container-name", method="GET")\n- NEVER attempt to fetch logs from a multi-container pod without specifying the container name\n- Examples of user requests that specify containers:\n  - "get logs from pod-name container nginx" → kubernetes_api_request(url="/api/v1/namespaces/default/pods/pod-name/log?container=nginx", method="GET")\n  - "show logs for container sidecar in pod-name" → kubernetes_api_request(url="/api/v1/namespaces/default/pods/pod-name/log?container=sidecar", method="GET")\n- If you encounter an error about "container name must be specified", the error handler will automatically provide guidance to the user\n- IMPORTANT: Parse user requests carefully to detect if they\'re specifying a container name in their request:\n  - Look for patterns like "container [name]", "from container [name]", "[pod-name] [container-name]"\n  - If user specifies a container name, use it directly in the log URL\n  - If user doesn\'t specify a container and the pod has multiple containers, fetch pod details first to list available containers' is too long - 'tools.0.function.description'",
       "type": "invalid_request_error",
       "param": null,
       "code": null
   }
}