Server providing Argo CD integration for MCP clients.
Nothing to run from here
Deskhand is installed from its publisher's own source and answers where it runs, so this marketplace is not in the path of a single call. There is no address here to send one to, and a panel that pretended otherwise would be showing you an answer we made up. Install it and call it from your own client — the Installation tab has the entry for each one.
93/100
trust score
12
tools declared
local
runs on your machine
—
calls last 7d
What it does
Safe KDE Plasma (Wayland) desktop control for AI agents: OCR-checked input, approval panel.
Quickstart
Install it the way its author documents:
1. Download deskhand-mcp.mcpb and open it in Claude Desktop or another client that installs MCP bundles (.mcpb)
2. Its sha256 is 8299a463df0e2761c12707b4e4faab5a28717eef1a2a922ea13949f01a074c09
The full guide: https://github.com/sobelendimm/deskhand/releases/download/v0.5.1/deskhand-mcp.mcpb
Publishernot claimed
SOAll 1 server
sobelendimm1 server
1 category
Collected from a public index. Nobody has claimed this account, so nothing here was written by its author — claim it if it is yours.
Installation
Where are you running it?
Every route below installs the same thing and ends at the same approval screen. This one runs on your machine: your client starts Deskhand as a process under your own user, with your files and your network, so the tool surface below is what it can do to your computer rather than to a server somewhere else. Its scan result is on the Security tab — read the surface before you approve it.
Install it the way its author documents:
1. Download deskhand-mcp.mcpb and open it in Claude Desktop or another client that installs MCP bundles (.mcpb)
2. Its sha256 is 8299a463df0e2761c12707b4e4faab5a28717eef1a2a922ea13949f01a074c09
The full guide: https://github.com/sobelendimm/deskhand/releases/download/v0.5.1/deskhand-mcp.mcpb
Tool surface
What the model actually sees. Descriptions are diffed on every release — see version history.
Read from its source code. A server that builds its tool list when it starts can list more tools than this.
Call FIRST. Stop switch, whether the user is busy (fullscreen video, game, recent input), the session's desktop choice, the background desktop state; windows=true lists windows (auto = both desktops).
| Parameter | Type | Required | Notes |
|---|---|---|---|
| windows | boolean | no | Also list windows |
| desktop | no | — |
Ask the USER where to do the task when they are busy (notification: Background / On my screen / Later; waits up to ~45 s). Remembered for this session (~10 min). choice "later" = do not do it now, tell the user.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| task | string | yes | Short task summary shown to the user (max 120 chars) |
Start/stop/inspect the invisible background desktop (Xvfb). stop closes every window on it.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| action | string | yes | — |
| size | string | no | WIDTHxHEIGHT for start, e.g. 1920x1080 |
Image of a window (preferred: fewer tokens) or the whole screen. Coordinates are pixels of THIS image; pass its shot_id to click/scroll with x,y. Forbidden windows are blacked out.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| window | no | — | |
| max_edge | integer | no | Long edge of the returned image, px |
| region | array | no | [x, y, w, h] zoom; in shot_id image pixels if shot_id given, else window pixels |
| shot_id | string | no | Earlier shot the region refers to |
| format | string | no | — |
| desktop | no | — |
Read the text in a window by OCR (Turkish+English) or the accessibility tree. Boxes/centers are WINDOW pixels;
| Parameter | Type | Required | Notes |
|---|---|---|---|
| window | no | — | |
| query | string | no | Only return matches of this text (Turkish-aware, fuzzy) |
| source | string | no | — |
| region | array | no | [x,y,w,h] window px |
| wait_s | number | no | With query: wait until it appears |
| deep | boolean | no | Slower multi-pass OCR (coloured buttons) |
| max_items | integer | no | — |
| desktop | no | — |
Find a saved icon (template match) or coloured areas; does not click. For text use read_text. Centers are window pixels (use the returned shot_id to click them).
| Parameter | Type | Required | Notes |
|---|---|---|---|
| kind | string | yes | — |
| window | no | — | |
| name | string | no | Saved icon name (kind=icon) |
| color | string | no | yesil|kirmizi|mavi or R,G,B (kind=color) |
| threshold | number | no | — |
| region | array | no | — |
| desktop | no | — |
Click visible text (preferred), an icon, an accessibility element, or x,y of a screenshot (needs its shot_id;
| Parameter | Type | Required | Notes |
|---|---|---|---|
| window | string | no | Target window (title pattern or id); may be omitted with shot_id |
| text | string | no | Visible text to click (preferred) |
| icon | string | no | Saved icon name |
| element | string | no | Accessibility element name |
| x | number | no | Image x of shot_id |
| y | number | no | Image y of shot_id |
| shot_id | string | no | — |
| nth | integer | no | Which match when the text appears several times |
| button | string | no | — |
| offset | array | no | [dx,dy] from the text centre, e.g. field next to a label |
| wait_s | number | no | Wait for the text to appear |
| scroll_search | boolean | no | Scroll down to find the text |
| verify | boolean | no | Report whether the window changed |
| desktop | no | — |
Type into the focused field of the window (click the field first). Stops if the window/title changes or the user starts typing. Enter on a payment page needs the user's approval.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| window | yes | — | |
| text | string | yes | Text to type (Turkish letters OK; = Enter; on a payment page Enter/Tab/space need the user's approval) |
| submit | boolean | no | Press Enter afterwards |
| clear_first | boolean | no | Select all (ctrl+a) first so the text replaces it |
| secret | boolean | no | Do not write the text to the log |
| desktop | no | — |
Send a key or shortcut to the window. Held keys are always released (cancel, STOP, shutdown). Enter/Space on a payment page needs the user's approval.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| window | yes | — | |
| keys | string | yes | Key or shortcut: enter, esc, ctrl+s, alt+F4, 'ctrl+l ctrl+c' (sequence) |
| repeat | integer | no | — |
| hold_s | number | no | Hold the key(s) down this long (games) |
| desktop | no | — |
Scroll the window (at its centre, or at x,y of shot_id). Returns whether the view changed.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| window | yes | — | |
| direction | string | no | — |
| amount | integer | no | — |
| x | number | no | — |
| y | number | no | — |
| shot_id | string | no | Needed with x,y (image pixels of that shot) |
| desktop | no | — |
Start an application on the chosen desktop and wait for its new window (returns its id). Forbidden apps are not started. On the background desktop the user's logged-in sessions are reused when the app's policy allows (copy/share); single-instance apps already open on the user's screen may be…
| Parameter | Type | Required | Notes |
|---|---|---|---|
| app | string | yes | Application name (.desktop entry) or executable |
| args | array | no | Arguments (runs app as an executable) |
| title | string | no | Wait for a window whose title matches this regex |
| wait_s | number | no | — |
| desktop | no | — |
List or run a saved flow (screen recognition -> steps, until success). Every step passes the same safety checks; a sensitive step stops the flow. Flows that contain pre-approved steps need the user's approval.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| action | string | no | — |
| name | string | no | Flow name (from list) |
| window | string | no | Target window for run |
| timeout_s | number | no | — |
| desktop | no | — |
Free
$0 forever
- Every tool, no call limit
- No card, and mcprush charges nothing for it
- Source published under GPL-3.0
- Runs on your machine — nothing of it reaches our gateway
- Nothing to cap, because mcprush bills nothing
What counts against your monthly calls
| Tool | Unit | Calls used | Out of the allowance |
|---|
Nothing here is billable: mcprush charges nothing to install Deskhand or to call it, at any volume.
Call allowance · this installenforced at the gateway
Monthly allowance, in calls
—
—
Alert at 80% · then the tool returns a normal MCP error
Used this month: —
ATrust score 93 / 100
Release 0.5.1 · scanned 8 Nov 2026
Two independent axes, because powerful and malicious are different questions. The grade is threat only. The capability level is blast radius, and it is never a penalty on the grade — it is priced as one subtract-only term in the score, where you can see it.
Threat score100
Client score93
Findings11 threat · 0 capability
Capabilityhigh
Coveragesource
Verificationrepo
Surface digest7e91f440b805e75e2da64918b2361ccf6ec4d37d3a487bad4d6f47b76182ecde
Score vectorthreat 100 → client 93
| Term | Level | What it prices | Points |
|---|---|---|---|
| capability-exposure | high | capability blast radius (high) — client exposure if the model is manipulated | −6 |
| verification-discount | repo | publisher verification (public source) — no provenance, but the source is public and inspectable | −1 |
| coverage-honesty | source | inspection depth (source) — how much of the target the scan could see | −0 |
score = clamp(round(threatScore − E_cap − E_ver − E_cov), 0, 100). No model, no opinion, no network call at scoring time: the same methodology version against the same bytes always produces this number, and every line above is a field of the report rather than a summary of it.
What the scan could actually read
A grade is only as meaningful as its coverage, so the scanner publishes its own depth before it publishes its result.
Tool surfaceread
Implementation sourceread
Package metadataread
Live transportnot available
Tools were statically extracted from the published source (12 recovered), not enumerated from a running server. Tool-poisoning, Unicode-smuggling, capability and toxic-flow analysis ran on this inferred surface, but a mis-parsed registration could be missed or mis-attributed, so tool-derived findings are capped below “confirmed”. To grade the real runtime surface, scan the running server: --command "npx -y <package>".
Capability — what it could do if the model were manipulated
Tags derived from each tool’s schema and the implementation, not from what the tool calls itself. high is the level these add up to.
| Tool | Capability tags | Why the tag was assigned |
|---|---|---|
| get_status | no tags | |
| choose_desktop | no tags | |
| manage_desktop | no tags | |
| screenshot | no tags | |
| read_text | no tags | |
| find | no tags | |
| click | no tags | |
| type_text | no tags | |
| press_key | no tags | |
| scroll | no tags | |
| open_app | no tags | |
| run_flow | no tags |
Toxic-flow graph
The lethal trifecta, checked as a graph rather than as a checklist: untrusted input, a sensitive source and an external sink have to meet before there is a path worth worrying about.
The public result for this release does not print the flow graph, so there is nothing to show here. That is not the same as "no paths were found": what the scan did read is above, under coverage.
Supply chain and provenance
Integrityfirst-seen
Surface digest7e91f440b805e75e
Scanned2026-11-08
This is the first scan of this surface here, so there is nothing yet to compare it against.
Every result on this tab comes from one deterministic pass over the published package — offline, rule by rule, and auditable line by line above. Same methodology version, same bytes, same score.
Release history
It installs from its repository rather than from a versioned package, so the command cannot be pinned to a release. The history below is for reading; the current release on file is 0.5.1.
0.5.1
undated
latestNo release note was published with this version.
—
no reviews yet
Who can review
One review per account, from a signed-in account that does not publish this listing. Nothing else is asked — you can say what you think before you install it. Publishers can reply once per review, and a review can be edited or deleted by whoever wrote it.
Sign in to write one
Any signed-in account that does not publish Deskhand can review it, once — before installing it or after. A review from an account that has installed it is marked as one.
No reviews yet
Nobody has reviewed this listing. The rating on the card is the mean of the reviews written here and nothing else, so there is no rating until somebody writes the first — which takes a signed-in account that does not publish it, and nothing else.