Desktop Commander MCP
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., "@Desktop Commander MCPlist the files in my Downloads 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.
desktop_commander
One capability bit from GBT V8, packaged standalone. No. 293 of 341,
tentacle domain (Tentacle).
桌面指挥官(Desktop Commander MCP):终端+文件系统全能力触手;支持本地 stdio 与 remote 设备模式
That description is copied verbatim from the registry. tests/test_spec.py pins it,
so documentation drift fails the build instead of shipping quietly.
What this package actually is
It is the contract for one capability: its name, its parameters, the values it accepts, and whether it can change state that cannot be restored. It is not the implementation — that lives in the runtime this bit was extracted from.
So run() goes through a narrow Backend protocol, and the default backend is a
stub that says so:
status="validated" backend="stub" data={"executed": false}Never status="ok". A stub that reports success is worse than no stub, because the
caller stops looking for the real thing.
Related MCP server: remotessh-mcp
Install
git clone https://github.com/gbtweilai-star/gbt-desktop-commander.git
cd gbt-desktop-commander
python -m pip install -e ".[dev]"Use
python -m gbt_desktop_commander info # capability card
python -m gbt_desktop_commander spec --json # machine-readable contract
python -m gbt_desktop_commander check --payload '{}'
python -m gbt_desktop_commander run --payload '{}'Real info output from this machine:
desktop_commander (tentacle) v1.0.0
----------------------------------------
桌面指挥官(Desktop Commander MCP):终端+文件系统全能力触手;支持本地 stdio 与 remote 设备模式
gated : False
external : False
multi-op : False
params : (none)Wire a real implementation:
from gbt_desktop_commander import run
class MyBackend:
name = "my-backend"
def invoke(self, cap, payload):
return {"cap": cap, "handled": sorted(payload)}
result = run({}, backend=MyBackend())
assert result.status == "ok"Parameters
name | required | notes |
— | no | the registry declares no required parameters |
Registry hint (params_hint): (empty)
Types are deliberately absent: the registry does not declare them, and an invented type is worse than no type, because callers believe it.
Boundaries
Not gated: the registry marks this bit
safe: true.No external service dependency declared.
Single-action bit.
No implementation ships here. Bring a backend; the protocol is one method.
What this is not
Not a wrapper around a hosted API. Not a model. Not a promise that calling it does anything on its own. It is the typed seam plus the validation, which is the part that stays correct even when the implementation changes.
Tests
python -m pytest20 passed, 3 skipped in 0.14sProvenance
field | value |
capability |
|
domain |
|
handler |
|
safe |
|
external |
|
atlas page | |
full index |
The parent README says 348 capabilities; an AST pass over the registry enumerates
341 + 0 unimplemented + 0 disabled = 341. The extra 7 are historical duplicates.
Numbering here follows 341 and matches the gbt-caps index one-to-one.
License
MIT.
This server cannot be deployed
Maintenance
Related MCP Connectors
Machine-native capabilities with explicit contracts and machine-readable commerce.
Provides capabilities that let LLM agents perform a range of infrastructure management tasks.
Operate Linux, macOS and Windows from your LLM. Every action runs through an auditable allowlist.
Drive real devices from your AI Coding tool. Embed a client SDK (Unity, Godot, Flutter, iOS/macOS, Android, React Native, Web) in your app, then capture screenshots, traverse the UI tree, inject taps and key events, and run automated test tasks on the physical device over a secure relay.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables secure terminal command execution, directory navigation, and file system operations through a standardized interface.10MIT
- AlicenseNot gradedqualityCmaintenanceProvides local command execution, remote SSH, interactive terminals, file read/write, and source search for AI CLI through stdio, with large output pagination and safety confirmations.5 npm2Apache 2.0
- AlicenseNot gradedqualityBmaintenanceEnables AI agents or human operators to spawn, monitor, drive, and hook into child terminal sessions—such as CLI tools, interactive shells, and polyglot runtimes—via POSIX pseudo-terminals, providing multiplexed control, event interception, and memory-bounded observability for orchestration.MIT
- AlicenseAqualityBmaintenanceEnables an AI agent to run local tools on the user's own machine via stdio, including command execution, workspace file read/write, and system status checks.52MIT