Bug Report: Local LLM Unable to Access File System via MCP in Void Editor

Type: Bug

I've tried using a MCP to overcome the problem of a local LLM being unable to access a file system because I haven't been able to achieve that in any other way. From what I can tell this is a widespread problem looking in the questions in discord there are many people suffering with the same issue.

Void Editor Version: [1.4.4]

Operating System: [ Windows 10]

Problem Description: I am unable to get a local Large Language Model (LLM) running via Ollama to access or modify my file system within Void Editor's Agent Mode, despite correctly configuring an MCP File System server. The LLM consistently states it cannot access the local file system.

Steps to Reproduce:

Ensured Docker Desktop is running and configured. Configured mcp.json in Void Editor to connect to a Docker-based MCP File System server. The content of mcp.json is: JSON

{ "mcpServers": { "filesystem": { "command": "docker", "args": [ "run", "-i", "--rm", "-v", "D:/story-buddy:/local-directory", "mcp/filesystem", "/local-directory" ] } } } (Note: D:/story-buddy is the actual local project directory, adjusted for JSON path compatibility). Configured Void Editor to use Ollama as a Local Provider and selected a local Ollama model (hf.co/tensorblock/gemma-3-12b-it-GGUF:Q8_0) in the "Models" settings. Confirmed the filesystem MCP server is enabled in Void's MCP settings, and the command displayed matches the mcp.json configuration. Attempted to prompt the LLM in Agent Mode with commands such as: "Please see if you can create a text file called ABC in the Story-buddy root directory and then add the word hello in the file." and "Using the file system tools, create a file named test.txt in the current directory and write 'Hello from LLM' into it." Observed the LLM's response: The LLM consistently replies that it cannot access or modify the file system and operates in a sandboxed environment, suggesting it's behaving like a cloud-based model without local access. (e.g., "As a large language model, I cannot directly access or modify your file system... I exist solely within Google's servers.") Expected Behavior: The local LLM, operating through Void Editor's Agent Mode and utilizing the configured Docker MCP File System server, should be able to perform file operations (read, write, create) within the specified mounted directory (D:/story-buddy).

Observed Errors/Warnings in Void Editor's Developer Tools Console:

The following critical error is consistently present in the Developer Tools console:

ERR Extension 'ms-python.vscode-pylance' appears in product.json but enables LESS API proposals than the extension wants. package.json (LOSES): mcpConfigurationProvider product.json (WINS): [email protected]:35 Note: This error persists even after disabling the Pylance extension. It suggests a fundamental conflict or misconfiguration within Void Editor's core API management (product.json vs. package.json) preventing the mcpConfigurationProvider from being properly exposed. Additionally, the following warnings/errors are present, though they appear to be less directly related to the core MCP issue:

workbench.desktop.main.js:4869 [Extension Host] Git installation not found. WARN Via 'product.json#extensionEnabledApiProposals' extension 'ms-python.python' wants API proposal 'terminalShellType' but that proposal DOES NOT EXIST. Likely, the proposal has been finalized (check 'vscode.d.ts') or was abandoned. (and similar for ms-python.vscode-python-envs and github.copilot-chat) workbench.desktop.main.js:1040 InitializeModel error: Hp: Unable to read file 'd:\story-buddy.voidrules' (Error: Unable to resolve nonexistent file 'd:\story-buddy.voidrules') Diagnostic Actions Taken:

Verified Docker Desktop is running. Confirmed mcp.json syntax and path (using forward slashes). Restarted Void Editor multiple times after configuration changes. Disabled the ms-python.vscode-pylance extension to isolate the specific error, but the mcpConfigurationProvider error persisted. Confirmed the local Ollama LLM is correctly selected and being used in the chat. Request for Assistance: Please investigate the mcpConfigurationProvider (LOSES) error as it appears to be blocking the Agent Mode from successfully utilizing configured MCP tools for local file system access.

Void version: Void 1.99.30039 (da425ab0fe34edfc3c082367edff7e05e0719b2d, 2025-06-10T00:36:08.103Z) OS version: Windows_NT x64 10.0.19045 Modes:

System Info

Item	Value
CPUs	Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz (8 x 3600)
GPU Status	2d_canvas: enabled canvas_oop_rasterization: enabled_on direct_rendering_display_compositor: disabled_off_ok gpu_compositing: enabled multiple_raster_threads: enabled_on opengl: enabled_on rasterization: enabled raw_draw: disabled_off_ok skia_graphite: disabled_off video_decode: enabled video_encode: enabled vulkan: disabled_off webgl: enabled webgl2: enabled webgpu: enabled webnn: disabled_off
Load (avg)	undefined
Memory (System)	31.95GB (21.16GB free)
Process Argv
Screen Reader	no
VM	0%

Extensions (5)

Extension	Author (truncated)	Version
vscode-containers	ms-	2.0.3
vscode-docker	ms-	2.0.0
debugpy	ms-	2025.8.0
python	ms-	2025.6.1
powershell	ms-	2025.0.0