Dittu vs Desktop Touch MCP
A side-by-side look at the Dittu and Desktop Touch MCP MCP servers: transport, auth, tools, quality signals, and the config to install either one in Claude, Cursor, Windsurf or VS Code.
Verdict: choose Dittu for os automation work over a hosted endpoint; choose Desktop Touch MCP for os automation as a local process. You can also run both side by side in one mcpServers block.
Which MCP server should you choose?
Choose Dittu when
- You need dedicated capabilities in the OS Automation domain.
- You prefer a remote HTTP/SSE transport.
- You want tools like
take_screenshot, take_screenshot_with_grid, take_screenshot_with_crosshair.
Choose Desktop Touch MCP when
- You need dedicated capabilities in the OS Automation domain.
- You prefer a local stdio subprocess transport.
- Its auth model fits you: No auth required.
- You have the keys it needs:
DESKTOP_TOUCH_MCP_HOME, GITHUB_TOKEN, GH_TOKEN, DESKTOP_TOUCH_MCP_ALLOW_UNVERIFIED. - You want tools like
desktop_discover, desktop_act, desktop_state.
Feature & specification comparison
Rows where the two servers differ are marked.
| Spec | ||
|---|---|---|
| Category | OS Automation | OS Automation |
| Quality signal | 53/100 · Good | 65/100 · Great |
| Install command | Confirmed | Confirmed |
| Transport | Remote HTTP/SSE | Local subprocess (stdio) |
| Auth | Not declared | No auth required |
| Pricing | Not declared | Free |
| Required env vars | None listed | DESKTOP_TOUCH_MCP_HOMEGITHUB_TOKENGH_TOKENDESKTOP_TOUCH_MCP_ALLOW_UNVERIFIED |
| Verified owner | Community listing | Community listing |
| GitHub stars | 0 | 22 |
| Directory activity | 0 views 0 installs 0 upvotes | 2 views 0 installs 0 upvotes |
| Tools | 12 | 23 |
| Full listing | View Dittu | View Desktop Touch MCP |
Tools & capabilities
Dittu tools 12
take_screenshotCapture the current screen via the Dittu phone camera. Returns a JPEG image. Call this to see what's on screen before acting.take_screenshot_with_gridCapture the screen and overlay a coordinate grid. step: grid spacing (default 0.1, use 0.05 for finer grid on small UI elements). Returns a JPEG image with grid drawn on top.take_screenshot_with_crosshairCapture the screen, crop tightly around (nx, ny), zoom in, and draw a crosshair at exactly that expected position. Use this for screen-corner calibration: move_mouse(nx, ny) near a screen edge, then call this to see whether the actual cursor lines up with the crosshair (where it SHOULD be if calibration is perfect). If the cursor is offset from the crosshair, or missing entirely (edge is cropped out of frame), corners need adjusting — see get_corners/set_corners. nx, ny: the same coordinates you passed to move_mouse. zoom: magnification factor for the cropped region (default 4x).get_raw_screenshotCapture the RAW camera frame — no perspective warp, no crop to the calibrated screen quad. Use this to see exactly what the camera sees before any correction, e.g. when troubleshooting why take_screenshot looks wrong, or before adjusting corners with get_corners/set_corners.get_cornersGet the phone's current screen-corner calibration. Returns {"tl": [nx,ny], "tr": [nx,ny], "br": [nx,ny], "bl": [nx,ny]} — each corner is a normalized point (0.0–1.0) in the RAW camera frame, defining the perspective warp that maps the camera view onto a flat rectangle representing the monitor. These four points are an arbitrary quadrilateral (trapezoid), not a bounding box — adjust one corner at a time without assuming the others move symmetrically.set_cornersUpdate the phone's screen-corner calibration. Each corner is [nx, ny] normalized 0.0–1.0 in the raw camera frame. tl/tr/br/bl = top-left/top-right/bottom-right/bottom-left of the monitor as seen by the camera (an arbitrary trapezoid due to perspective — not a rectangle). Call get_corners() first to read the current values before nudging one of them.move_mouseMove mouse cursor to normalized coordinates WITHOUT clicking. nx, ny: float 0.0000–1.0000.clickClick at normalized screen coordinates. nx, ny: float 0.0000–1.0000 where (0,0) = top-left, (1,1) = bottom-right.double_clickDouble-click at normalized screen coordinates.type_textType text on the computer keyboard.keyPress a key combination. modifiers bitmask: LCtrl=1, LShift=2, LAlt=4, LWin=8 Common keycodes: Enter=40, Escape=41, Tab=43, Space=44, F4=61, PageDown=78, PageUp=75, Win=227 Examples: Alt+F4 → key(4, [61]) Ctrl+W → key(1, [26])scrollScroll the mouse wheel. delta: number of ticks. Positive = scroll down, negative = scroll up.
Desktop Touch MCP tools 23
desktop_discoverObserve the desktop. Returns interactive entities with leases (UIA, CDP, Terminal, Visual SoM).desktop_actPerform actions (click, type, set a value, select) on entities via lease validation. Returns semantic diffs — plus an optional `roiCapture` (changed-region PNG + next-target preview) on visual-only targets.desktop_stateLightweight check of focus, active window, cursor, and Auto-Perception attention signal.screenshotMulti-mode capture: `detail='text'` (UIA/OCR), `diffMode` (P-frame), `dotByDot` (1:1), and `background`. Returns a cheap `screenshot://by-ref/{id}` link to the saved image instead of inlining pixels every time.screenshot_queryInspect and prune the on-disk screenshot cache behind the by-ref links: `screenshot_query` lists saved captures without re-reading pixels; `screenshot_gc` reclaims space by retention policy (dry-run by default).workspace_snapshotInstant session orientation: all window thumbnails + UI summaries in one call.server_statusDiagnostic check for native engine health and feature activation.keyboardSend keyboard input. Supports background input (WM_CHAR) and IME-safe clipboard bypass.mouse_clickPrecision coordinate-based interaction with homing and force-focus protection.scrollMulti-strategy: `raw` (notches), `to_element`, `smart` (virtual lists), and `capture` (stitch).click_elementLegacy UIA-based click by name/ID (fallback when entities are unavailable).browser_openIdempotent debug-mode launch and reliable navigation.- +11 more on the listing page
Ready-to-paste client configs
Paste either block (or both) into your client config, e.g. claude_desktop_config.json or ~/.cursor/mcp.json. We only show a config when the install command is confirmed.
Dittu
{
"mcpServers": {
"dittu": {
"url": "https://app.dittu.org/mcp"
}
}
}
Desktop Touch MCP
{
"mcpServers": {
"harusame64-desktop-touch-mcp": {
"command": "npx",
"args": [
"-y",
"@harusame64/desktop-touch-mcp"
],
"env": {
"DESKTOP_TOUCH_MCP_HOME": "YOUR_DESKTOP_TOUCH_MCP_HOME_HERE",
"GITHUB_TOKEN": "YOUR_GITHUB_TOKEN_HERE",
"GH_TOKEN": "YOUR_GH_TOKEN_HERE",
"DESKTOP_TOUCH_MCP_ALLOW_UNVERIFIED": "YOUR_DESKTOP_TOUCH_MCP_ALLOW_UNVERIFIED_HERE"
}
}
}
}
Frequently asked questions
Dittu is categorized under OS Automation and uses a remote HTTP/SSE transport. Desktop Touch MCP belongs to OS Automation and uses a local stdio subprocess. Pick Dittu for os automation work and Desktop Touch MCP for os automation.
Copy the mcpServers block from the config section on this page into your client's config (for example claude_desktop_config.json or ~/.cursor/mcp.json), then fully restart the client.
Dittu has no declared pricing — check the project's own site. Desktop Touch MCP is listed as free. Auth: no auth required. If a server needs environment variables such as API keys, define them under the env key in your client's MCP config.
Yes. MCP clients support multiple servers: add dittu and harusame64-desktop-touch-mcp as separate keys in the same mcpServers object, and your assistant will call tools from either one.
On AllMCPs, Dittu has a quality signal of 53/100 (Good) with 0 views, 0 installs and 0 GitHub stars. Desktop Touch MCP has 65/100 (Great) with 2 views, 0 installs and 22 GitHub stars.