Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@ownerops-desktop-mcp list the files in my Documents folder"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Self-hosted personal desktop MCP. The goal is a small, inspectable alternative to a paid desktop-control bridge rather than a feature-for-feature clone.
get_policy
list_directory
read_file
write_file
search_files
run_shell
No delete, registry, privilege-escalation, or arbitrary process-kill tool is exposed.
Related MCP server: Windows Agent MCP Server
Filesystem access is constrained by DESKTOP_MCP_ROOTS. Every file path is resolved before
use and must remain under an allowed root. Sensitive directory names can additionally be denied
with DESKTOP_MCP_DENY_NAMES.
The shell is deliberately restricted:
its working directory must be inside an allowed root
only commands in DESKTOP_MCP_SHELL_ALLOW may start
shell chaining and redirection are rejected
runtime and output are capped
The shell policy is not an OS sandbox. An allowlisted executable may still reach host resources. For a hard boundary, run under a dedicated low-privilege account or add a container sandbox.
HTTP mode supports:
oauth — OAuth authorization code flow, S256 PKCE, dynamic client registration,
refresh-token rotation, and offline_access
bearer — static bearer token for simple MCP clients
none — local/private testing only
OAuth credentials and tokens live outside Git in .env.
git clone https://github.com/sionchu/ownerops-desktop-mcp.git
cd ownerops-desktop-mcp
copy .env.example .env
uv sync --dev
uv run pytest
uv run ruff check .Set explicit roots and generate strong OAuth credentials before remote use.
uv run python http_server.pyLocal MCP endpoint: http://127.0.0.1:8765/mcp
start_mcp.cmd is provided for Windows startup/task-scheduler use.
If the public base is https://host.example/ownerops:
tailscale funnel --bg --set-path /ownerops http://127.0.0.1:8765RFC 9728 protected-resource metadata is advertised at a root well-known URL. If another
service already owns /, add a narrow route for that exact metadata path as well.
Prefer a private Tailscale path or OpenAI Secure MCP Tunnel when available instead of a public Funnel. If Funnel is used, keep OAuth enabled.
The project tracks MCP Python SDK 2.x and CI runs on Windows with pytest and Ruff.
MIT
This server cannot be deployed
Hosted MCP catalog with 30 tenant-isolated browser, RAG, AI, mail and media tools.
MCP gateway with runtime security policy, tool-call-level control, and audit of agent actions.
MCP server for mandates, delegation, policy-gated execution, credential grants, and audit.
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
