WinHelm | Monitor Dashboard

⚡ WinHelm

LIVE MONITOR PROFILE: POWER · MODIFIED · 41 SCHEMAS

Auto-refresh (2.5s)

Legacy configuration — upgrade available. This config has no schemaVersion, so v1.2.1 permissions, profile filtering and allowed directories stay in effect. Approval-gated policies (such as approval for terminating processes WinHelm did not start) are schema-v2 features and are not active in legacy mode. Review the schema-v2 upgrade in Permissions with a local admin code.

Permissions & Setup Locked · local admin code required

Run winhelm admin-code in a local terminal, then enter the one-time code here. The admin credential is separate from MCP authentication. Never paste the code into an AI chat.

Profile Wizard

Profile is a preset only. The permission state shown below is authoritative.

Preset Folders · one per line, optional path::alias Typed confirmation · Power/full only

Unlock completed. Loading current permissions…

MCP Client Connection

Writes only mcpServers.winhelm in the local client config. Unrelated settings are preserved, an existing file is backed up first, and no auth or admin secret is embedded.

No client preview yet.

Approval Jobs

Pending approvals are local-only. Arguments are displayed as text, never rendered as HTML/Markdown.

No admin data loaded yet.

Connected MCP Clients

Client names are supplied by the connecting client and are displayed as text only.

No session has announced clientInfo yet.

MCP Access Token

Rotating replaces the bearer token immediately: clients still using the old token fail on their next request. Dashboard session cookies issued earlier stay valid for up to 15 minutes.

—

Test and teach your AI

Run one of these prompts in your AI app, then watch the Activity & Execution Stream further down this page until the first successful call appears.

Example prompts

Use WinHelm to list the files in my workspace folder.

Use WinHelm to read the first 20 lines of a log file on my PC.

Instructions for the AI

When I mention files, apps or processes on my computer, use the WinHelm tools. They run on my real Windows PC, not a sandbox. If a tool result contains userAction, tell me what it says instead of working around it.

Claude Desktop can answer from its own sandbox instead of WinHelm. In one measured run, 0 of 6 natural prompts reached WinHelm, while an explicit prompt containing "ใช้ mcp" reached it 1 of 1 with the full schema. Adding "ใช้ mcp" or "ผ่าน WinHelm" is worth trying; it is not a guarantee.

If the AI still answers from a sandbox, try turning off "Code execution and file creation" in that app's settings. This is an observation from testing, not a guarantee.

Server Runtime

WinHelm v1.0.0

Connected Sessions

Streamable: 0 | SSE: 0 | Tasks: 0

CPU Load

-- %

Memory (RAM)

-- / -- GB

Storage Drives

Reading drive metrics...

GPU Telemetry

Detecting GPU...

Activity & Execution Stream

Time	Type	Event	Details	Latency	Status
No logs recorded yet

Exposed Tool Schemas (41 Exposed / 41 Available)

Terminal Background Tasks Filesystem Search & PDF Desktop System & Services Network

terminal_run

terminal_task_start

terminal_task_list

terminal_task_logs

terminal_task_send

terminal_task_kill

file_read

file_write

file_edit

file_list

file_search

file_delete_safe

file_move

file_copy

archive_zip

archive_unzip

file_tail

file_hash

file_search_ripgrep

pdf_generate

file_preview

clipboard_get

clipboard_set

screen_capture

system_open

notification_send

system_info

gpu_info

process_list

process_kill

port_check

eventlog_query

service_list

service_status

service_control

http_ping

http_request

network_info

Terminal adapter

command_run · terminal=full

Recovery tools

winhelm_status

winhelm_job_status

MCP resource

preview://file