ReMCP
Server Details
Securely control computers you explicitly pair through files, terminals, processes, screenshots, desktop UI/input, clipboard, browser automation, diagnostics, and document tools.
- Status
- Healthy
- Uptime
- 95.5% over the last 24 hours
- OAuth
- Works in Glama
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 178 tools
The surface contains many overlapping families: ui_click/computer_click/pointer_click, ui_invoke/computer_invoke, ui_set_value/computer_set_value, and browser_scroll_into_view/ui_scroll_into_view/computer_scroll_into_view all target similar actions. Descriptions add useful caveats, but the boundaries between the duplicated accessibility/desktop/browser families remain hard to infer reliably. Several descriptions also refer to browser_action or ui_action, which are not in the exposed tool list, increasing selection confusion.
Most names use snake_case and follow a verb_noun or resource_action pattern, but there are bare noun tools such as environment, notification, keyboard, and scroll, plus inconsistencies like take_screenshot versus screenshot_region. The many category prefixes are readable, though duplicate concepts are split across ui_, computer_, pointer_, and browser_ prefixes. Overall it is mixed but not chaotic.
At 178 tools this is far beyond a well-scoped surface, even for a broad remote-computer-control server. The extensive duplication among ui_, computer_, pointer_, app_, and browser_ operations makes the set feel accumulated rather than deliberately scoped. This is an extreme count mismatch.
The server covers an unusually wide lifecycle: device pairing, browser automation, native UI interaction, files, archives, documents, PDFs, spreadsheets, processes, services, network inspection, power control, audio, clipboard, and configuration/statistics. Major remote-control workflows have corresponding tools, and obvious dead ends are mostly absent. The main issue is overcoverage and duplication, not missing operations.
Publisher details
- Operator
- ReMCP · Publisher source
- Operator website
- https://remcp.site · Publisher source
- Vendor relationship
- First-party · Publisher source
- Documentation
- https://remcp.site/docs · Publisher source
- Privacy policy
- Not provided
- Trust center
- Not available
- Restrictions
- Requires signing in to ReMCP and pairing at least one device. Desktop and browser-control features depend on the paired device's OS permissions and advertised capabilities. · Publisher source
Available Tools
178 toolsapp_actionFill browser fields or perform a supported app taskADestructiveInspect
Use this when the user needs to fill 1–20 form fields in one selected Chromium tab or perform one supported application task on a paired computer. Browser form writes verify the exact origin, specified prior field values and resulting values. App tasks use the existing scoped permissions. Changes require commit=true and a unique operation_id UUID; otherwise the tool previews the action. Read-only inspections do not require commit. It does not automatically submit forms or run arbitrary scripts or commands.
Inputs: mode (required), device (required), task, commit, +7 more. Returns: data, text
app_catalogApplication GUI catalogARead-onlyIdempotentInspect
Use this when the user needs to discover curated workflows and, with include_detected=true, new OS software records. Linux prefers visible XDG GUI launchers; macOS .app copies have separate IDs and require PID executable provenance; Windows uses local Start Menu shortcuts with exact PID executable attestation; registry-only rows cannot authorize GUI access. Detection is not proof of an active GUI, permissions or a trusted connector. Use local- ids for inspect/find-visible-controls or explicit checked focus_control and one bounded set_control_text with old-value CAS; other effects are forbidden.
Inputs: device (required), app, filter, refresh, +4 more. Returns: data, text
apply_patchApply patchADestructiveInspect
Use this when you already have a unified diff with one or more hunks and want to apply it to one or several files. Do not use write_file for the same patch-shaped change, and prefer edit_block when the change is one exact text block rather than a diff.
Inputs: patch (required), device (required), path, dry_run. Returns: data, text
app_taskUnified application taskBDestructiveInspect
Use this when the user needs to execute one typed intent inside a curated app via trusted local API, scoped Accessibility or CDP without screenshots. New local- IDs support inspect/find-visible-controls and explicitly scoped non-destructive focus_control, after unique PID/window and semantic ID/name/role checks. Unknown apps cannot launch, click, invoke or submit; a single native text value can change only after explicit commit and exact old-value check (raw expected_text or an expiring blind token from inspect/find). Telegram edits require authenticated owner, original text and verified read-back.
Inputs: app (required), device (required), intent (required), pid, +14 more. Returns: data, text
audio_muteAudio MuteAInspect
Use this to mute the paired computer system output. It does not pause playback or record audio; check the existing mute state when avoiding unnecessary changes.
Inputs: device (required). Returns: data, text
audio_set_volumeAudio Set VolumeAInspect
Use this to set the paired computer system output volume to a requested supported level. This affects audible output immediately; it does not record or transmit audio.
Inputs: device (required), volume (required). Returns: data, text
audio_statusAudio StatusARead-onlyIdempotentInspect
Use this to inspect supported system output volume and mute state on a paired computer. It reads local audio settings without playing audio or changing the volume.
Inputs: device (required). Returns: data, text
audio_unmuteAudio UnmuteAInspect
Use this to unmute the paired computer system output on user request. It may make previously playing audio audible immediately; verify volume if necessary.
Inputs: device (required). Returns: data, text
browser_backBrowser BackADestructiveInspect
Use this to go back one page in the history of a selected Chromium tab. It may leave unsaved form data; choose the tab first and inspect the new page state afterward.
Inputs: device (required), wait, title, endpoint, +3 more. Returns: data, text
browser_clickBrowser ClickADestructiveInspect
Use this to click a verified element on the current Chromium page by CSS selector or visible text. Inspect the tab and target first; clicking may submit forms, purchase items or delete data.
Inputs: device (required), text, title, endpoint, +5 more. Returns: data, text
browser_evaluateEvaluate browser JavaScriptADestructiveInspect
Use this when browser_snapshot, browser_find, browser_action and browser_wait cannot express the required page operation or inspection and JavaScript is necessary. It is a powerful page-context escape hatch, and page content is untrusted.
Inputs: device (required), expression (required), title, endpoint, +4 more. Returns: data, text
browser_fill_formBrowser Fill FormADestructiveInspect
Use this to fill several verified form fields on one inspected Chromium page in a bounded operation. Validate the origin, selectors and supplied values; filling does not implicitly authorize form submission.
Inputs: device (required), fields (required), title, endpoint, +3 more. Returns: data, text
browser_findFind browser elementARead-onlyIdempotentInspect
Use this when the user needs to locate visible web elements by CSS selector, text or ARIA role and obtain reusable selectors/bounds. It requires an element criterion; use browser_snapshot, not an empty browser_find, to enumerate page structure. Prefer this before browser_action rather than guessing selectors or coordinates.
Inputs: device (required), role, text, limit, +6 more. Returns: data, text
browser_focusBrowser FocusAInspect
Use this to focus a specific existing Chromium page element by CSS selector or text before typing. It does not enter content; prefer browser_find for an unknown target.
Inputs: device (required), text, title, endpoint, +5 more. Returns: data, text
browser_forwardBrowser ForwardADestructiveInspect
Use this to move forward one step in the history of a selected Chromium tab. It requires previous backward navigation and may discard unsaved page state, even though it does not submit forms.
Inputs: device (required), wait, title, endpoint, +3 more. Returns: data, text
browser_new_tabBrowser New TabAInspect
Use this to open a specified URL in a new Chromium browser tab on the paired computer. This does not navigate an existing tab; the URL is subject to browser security restrictions.
Inputs: url (required), device (required), wait, title, +4 more. Returns: data, text
browser_pressBrowser PressADestructiveInspect
Use this to send an explicit keyboard key or shortcut to the current inspected Chromium tab, such as Enter or Escape. Enter can submit forms; validate focus and user intent first.
Inputs: key (required), device (required), title, endpoint, +3 more. Returns: data, text
browser_reloadBrowser ReloadADestructiveInspect
Use this to reload an identified Chromium tab, optionally bypassing cache. Reloading may lose unsaved form state; use a browser snapshot to verify the resulting page.
Inputs: device (required), wait, title, endpoint, +4 more. Returns: data, text
browser_scrollBrowser ScrollAInspect
Use this to scroll an identified Chromium tab by a bounded amount or direction. This changes the viewport only; use scroll_into_view when a known target element should become visible.
Inputs: device (required), delta, title, delta_x, +7 more. Returns: data, text
browser_scroll_into_viewBrowser Scroll Into ViewAInspect
Use this to scroll a specified Chromium page element into view using its verified selector or visible text. It does not click the element or change its value.
Inputs: device (required), text, title, endpoint, +5 more. Returns: data, text
browser_selectBrowser SelectADestructiveInspect
Use this to select an option from a native HTML select control on an inspected Chromium page. Verify the selector and valid option first; selection can trigger dependent form updates.
Inputs: device (required), text, title, option, +6 more. Returns: data, text
browser_set_valueBrowser Set ValueADestructiveInspect
Use this to set a specific Chromium form field value through guarded DOM interaction rather than individual keystrokes. Inspect the page and selector first; changing the field can trigger page events.
Inputs: device (required), text, title, value, +6 more. Returns: data, text
browser_set_viewportBrowser Set ViewportAInspect
Use this to change an inspected Chromium tab viewport width, height and optional mobile scale for responsive testing. It changes browser emulation, not the host monitor resolution.
Inputs: width (required), device (required), height (required), title, +6 more. Returns: data, text
browser_snapshotBrowser snapshotARead-onlyIdempotentInspect
Use this when the user needs to inspect a web page structurally through CDP accessibility. Reuse the last snapshot_id as diff_from to return changed/removed semantic nodes rather than retransmitting the whole page after each step; refresh with a full snapshot when unavailable. Prefer it over desktop ui_snapshot for page content; add include_screenshot=true only when final pixel/layout verification is needed. A selector may be temporarily centered for capture, but the original page scroll position is restored before the tool returns.
Inputs: device (required), title, endpoint, selector, +6 more. Returns: data, text
browser_tabsBrowser tabsARead-onlyIdempotentInspect
Use this when the user needs browser tabs: List debuggable Chromium tabs when CDP is known ready. If uncertain, first check computer_snapshot(browser_readiness=true, include_ui=false, include_screenshot=false, include_browser=false). When not ready, use native ui_snapshot/ui_find rather than retrying the unchanged endpoint. This read-only tool never launches a browser; use target_id/title/url returned for subsequent browser tools.
Inputs: device (required), title, endpoint, target_id, +2 more. Returns: data, text
browser_typeBrowser TypeADestructiveInspect
Use this to enter user-provided text into a verified Chromium page field, optionally clearing existing text. Confirm the origin and field first; do not put sensitive data into an untrusted page.
Inputs: text (required), device (required), clear, title, +6 more. Returns: data, text
browser_uploadBrowser UploadADestructiveInspect
Use this to attach a user-authorized local file to an inspected Chromium file-input element. Check the exact file path, site origin and target input first; the page may receive the file contents.
Inputs: device (required), selector (required), path, paths, +5 more. Returns: data, text
browser_waitWait for browserARead-onlyIdempotentInspect
Use this when browser navigation or an action must be followed by proof of selector/text/URL/load/navigation/network-idle instead of fixed sleeps or screenshot polling. Use browser_evaluate separately for JavaScript predicates.
Inputs: device (required), text, title, value, +8 more. Returns: data, text
browser_workflowBrowser WorkflowADestructiveInspect
Use this for an explicitly requested sequence of 1–12 guarded DOM actions and bounded waits on one inspected Chromium page. Validate CSS selectors first. The workflow cannot run arbitrary JavaScript, upload files, switch origins or implicitly submit a form; an explicit click may submit data or cause irreversible remote changes. Stop at the first failed step and never echo field values.
Inputs: steps (required), device (required), title, endpoint, +3 more. Returns: data, text
clipboard_clearClipboard ClearADestructiveInspect
Use this to clear the paired computer clipboard on user request, for example after handling sensitive copied text. This irreversibly removes its current contents from the clipboard.
Inputs: device (required). Returns: data, text
clipboard_readClipboard ReadARead-onlyIdempotentInspect
Use this when the user asks to inspect the current clipboard text on a paired computer. Clipboard contents may be sensitive; read them only when needed and do not send them to unrelated sites.
Inputs: device (required). Returns: data, text
clipboard_writeClipboard WriteADestructiveInspect
Use this to place user-authorized text on the paired computer clipboard for pasting into an application. This overwrites the previous clipboard contents but does not itself paste or submit data.
Inputs: text (required), device (required). Returns: data, text
computer_add_to_selectionComputer Add To SelectionADestructiveInspect
Use this to add an identified native list or tree item to an existing multi-selection without replacing the other selected items. Confirm the target and current selection first.
Inputs: device (required), x, y, id, +14 more. Returns: data, text
computer_clickComputer ClickADestructiveInspect
Use this to click an identified native desktop control or screen coordinate on a paired computer after inspecting the current UI. Clicking can submit forms, activate buttons or change data; use a semantic target when available.
Inputs: device (required), x, y, id, +14 more. Returns: data, text
computer_collapseComputer CollapseBDestructiveInspect
Use this to collapse an identified expanded tree item, menu or disclosure control. Inspect the target first and avoid hiding information needed for the next action.
Inputs: device (required), x, y, id, +13 more. Returns: data, text
computer_dragComputer DragADestructiveInspect
Use this for a real drag-and-drop or selection gesture between verified native desktop elements or screen coordinates. Confirm the source and destination first; dropping can move files or change application state.
Inputs: device (required), x, y, id, +22 more. Returns: data, text
computer_expandComputer ExpandADestructiveInspect
Use this to expand an identified native tree item, menu or disclosure control. This reveals child controls and may trigger application-side effects depending on the target.
Inputs: device (required), x, y, id, +13 more. Returns: data, text
computer_focusComputer FocusAInspect
Use this to bring an identified native desktop control into focus before typing or interacting with it. Prefer a current semantic target; focusing alone does not type text or activate a button.
Inputs: device (required), x, y, id, +13 more. Returns: data, text
computer_invokeComputer InvokeADestructiveInspect
Use this when the user asks to activate an identified native desktop accessibility control on a paired computer, such as invoking a menu item or button. Resolve the intended control from a recent computer_snapshot or ui_find first. Invoking it can submit forms, change application state, or trigger irreversible actions; it does not run arbitrary shell commands or page JavaScript.
Inputs: device (required), x, y, id, +13 more. Returns: data, text
computer_multi_editComputer Multi EditADestructiveInspect
Use this for a bounded batch of native accessibility edits on already identified desktop controls. Verify each requested change and target before applying; individual edits can change application data.
Inputs: edits (required), device (required), x, y, +14 more. Returns: data, text
computer_multi_selectComputer Multi SelectADestructiveInspect
Use this to select multiple verified native desktop items as one semantic accessibility operation. Supply identified targets from a fresh snapshot; selection alone does not open or delete files.
Inputs: device (required), targets (required), x, y, +14 more. Returns: data, text
computer_remove_from_selectionComputer Remove From SelectionADestructiveInspect
Use this to remove an identified native list or tree item from the current multi-selection while preserving remaining selections. It does not delete the underlying item.
Inputs: device (required), x, y, id, +14 more. Returns: data, text
computer_scrollComputer ScrollAInspect
Use this to scroll the native desktop viewport or an identified UI region by a bounded direction or delta. It moves the visible viewport and does not edit document contents.
Inputs: device (required), x, y, id, +18 more. Returns: data, text
computer_scroll_into_viewComputer Scroll Into ViewBInspect
Use this to bring an identified native accessibility element into the visible viewport before reading or interacting with it. This scrolls the interface without editing its data.
Inputs: device (required), x, y, id, +13 more. Returns: data, text
computer_selectComputer SelectBDestructiveInspect
Use this to select an identified option or item in a native desktop accessibility control. Inspect the control and its available values first; selection may update application state.
Inputs: device (required), x, y, id, +14 more. Returns: data, text
computer_set_range_valueComputer Set Range ValueBDestructiveInspect
Use this to set a numeric or range control, such as a slider, to a specific supported value on the paired desktop. Inspect the control limits first; changing it can affect application settings.
Inputs: device (required), x, y, id, +14 more. Returns: data, text
computer_set_valueComputer Set ValueADestructiveInspect
Use this to assign a specified value to an identified native accessibility control, such as an editable text field. Validate the target before writing; this changes the control rather than sending a keyboard shortcut.
Inputs: device (required), x, y, id, +14 more. Returns: data, text
computer_snapshotComputer snapshotARead-onlyIdempotentInspect
Use this when you need the first compact view of an unfamiliar desktop state, including windows, semantic UI, displays, clipboard metadata and optional pixels/browser/OCR. Prefer it over repeated screenshots; narrow later with ui_find, browser_find or screenshot_region.
Inputs: device (required), ocr_psm, ui_scope, grid_rows, +19 more. Returns: data, text
computer_toggleComputer ToggleADestructiveInspect
Use this to change the state of a native checkbox, switch or toggle identified through the current desktop snapshot. Check its current state to avoid unintended changes; toggles can alter application settings.
Inputs: device (required), x, y, id, +13 more. Returns: data, text
computer_typeComputer TypeADestructiveInspect
Use this to enter user-provided text into a focused or identified native desktop control, optionally clearing its previous value. First confirm the target from the current UI; never type passwords or private data into an unverified field.
Inputs: text (required), device (required), x, y, +15 more. Returns: data, text
copy_fileCopy fileADestructiveInspect
Use this when the user asks to copy exactly one regular file to a new path on the same computer. For directories or two or more copies use copy_paths.
Inputs: device (required), source (required), destination (required), overwrite. Returns: data, text
copy_pathsCopy pathsADestructiveInspect
Use this when the user asks to copy a directory recursively or two or more paths in one request. For exactly one regular file use copy_file.
Inputs: paths (required), device (required), overwrite. Returns: data, text
create_archiveCreate archiveADestructiveInspect
Use this when the user wants files or a directory packed into a tar, tar.gz, or zip archive, for example before a transfer or a backup.
Inputs: paths (required), device (required), destination (required), format. Returns: data, text
create_directoryCreate directoriesAIdempotentInspect
Use this when the user asks to create a directory, including any missing parent directories.
Inputs: device (required), path, paths. Returns: data, text
create_pairing_commandConnect a new computerAIdempotentInspect
Use this when the person wants to connect another computer to their ReMCP account, or asks how to add a device: returns platform-specific Windows/macOS/Linux commands, the pairing code, and when it expires. Windows uses the no-prerequisite PowerShell bootstrap; macOS/Linux use the npm launcher. The same commands are returned while the code is still valid.
Inputs: no named parameters. Returns: data, text
delete_pathDelete pathADestructiveInspect
Use this when the user explicitly asks to remove exactly one file or directory and does not need it recoverable. Use delete_paths for two or more targets, or move_to_trash when reversibility matters.
Inputs: path (required), device (required), recursive. Returns: data, text
delete_pathsDelete pathsADestructiveInspect
Use this when the user explicitly asks to remove two or more named files or directories in one batch. For one target use delete_path; use move_to_trash when reversibility matters.
Inputs: paths (required), device (required), recursive. Returns: data, text
diff_filesDiff filesARead-onlyIdempotentInspect
Use this when the user wants to see how two text files differ, before or after a change.
Inputs: left (required), right (required), device (required), context_lines. Returns: data, text
disable_deviceDisable paired computerADestructiveInspect
Use this when the user needs to disable an existing paired computer only when the user explicitly requests it. Disconnects the pairing until re-enabled; requires its device id.
Inputs: device (required). Returns: data, text
display_inventoryDisplay inventoryARead-onlyIdempotentInspect
Use this when monitor-specific screenshots, cross-monitor geometry or coordinate work require monitor bounds, scale and primary-display information.
Inputs: device (required). Returns: data, text
document_append_paragraphDocument Append ParagraphBDestructiveInspect
Use this to append a user-provided paragraph to a specified local word-processing document. Verify the path and creation option; this writes document content.
Inputs: path (required), text (required), device (required), create, +1 more. Returns: data, text
document_delete_paragraphDocument Delete ParagraphADestructiveInspect
Use this to delete one or more paragraphs matching requested text in a specified local word-processing document. This removes document content; confirm the exact match and backup needs.
Inputs: path (required), device (required), search (required), all, +2 more. Returns: data, text
document_insert_paragraph_afterDocument Insert Paragraph AfterADestructiveInspect
Use this to insert a new paragraph after an identified matching paragraph in a specified local document. Confirm the search target and intended location before writing.
Inputs: path (required), text (required), device (required), search (required), +2 more. Returns: data, text
document_insert_paragraph_beforeDocument Insert Paragraph BeforeADestructiveInspect
Use this to insert a new paragraph before an identified matching paragraph in a specified local document. Check the match and the text before writing.
Inputs: path (required), text (required), device (required), search (required), +2 more. Returns: data, text
document_prepend_paragraphDocument Prepend ParagraphBDestructiveInspect
Use this to insert a user-provided paragraph at the beginning of a specified local word-processing document. Confirm the document path and preserve existing content.
Inputs: path (required), text (required), device (required), create, +1 more. Returns: data, text
document_replaceDocument ReplaceADestructiveInspect
Use this to replace specified matching text in a user-selected local word-processing document. Check the search string, replacement and destination; edits change document content.
Inputs: path (required), device (required), search (required), replacement (required), +3 more. Returns: data, text
drag_dropDrag and dropADestructiveInspect
Use this when the task calls for a real drag gesture between two UI element ids or coordinates. Prefer element ids from ui_find/ui_snapshot because their current bounds are resolved at execution time; use coordinates only when no semantic endpoints exist. On GNOME Wayland, real drag input requires the consent-backed Remote Desktop portal because XTEST/xdotool drag delivery is not reliable.
Inputs: device (required), to_x, to_y, to_id, +7 more. Returns: data, text
edit_blockEdit fileADestructiveInspect
Use this when one known text block in one file should be replaced, with a whitespace-tolerant fallback and optional dry-run diff. Use write_file only when replacing the whole file, and replace_lines when the intended target is a known line range rather than matching text.
Inputs: device (required), file_path (required), new_string (required), old_string (required), +3 more. Returns: data, text
edit_spreadsheetEdit spreadsheetADestructiveInspect
Use this when the user needs to create a new XLSX workbook or directly edit XLSX cells/ranges/formulas without launching Excel. Set create=true only for a new path; it fails rather than overwriting an existing workbook. Prefer explicit cells/ranges and write to output when preserving an existing original matters.
Inputs: path (required), edits (required), device (required), sheet, +2 more. Returns: data, text
enable_deviceEnable paired computerAInspect
Use this when the user needs to enable an existing paired computer only when the user explicitly requests it. Changes that pairing to enabled; requires its device id.
Inputs: device (required). Returns: data, text
environmentRuntime environmentARead-onlyIdempotentInspect
Use this when the task calls for PATH, shell, Node/runtime environment paths, architecture and optionally sanitized environment variables. Use get_system_info for host health such as CPU/memory/disk/uptime, and get_runtime_info for ReMCP policy, roots and limits. Secret-looking values remain masked.
Inputs: device (required), include_env. Returns: data, text
event_logEvent logsARead-onlyIdempotentInspect
Use this when the user needs to read bounded recent OS logs (Windows Event Log, macOS unified log or systemd journal) when diagnosing a service/app/system problem. Prefer it over an unbounded shell log dump.
Inputs: device (required), log, limit, query, +2 more. Returns: data, text
extract_archiveExtract archiveADestructiveInspect
Use this when the user wants an existing tar or zip archive unpacked on a paired computer.
Inputs: device (required), archive (required), destination. Returns: data, text
force_terminateStop sessionADestructiveInspect
Use this when a process session created by start_process must stop, such as a runaway build or server left behind. For an arbitrary operating-system PID that is not a ReMCP terminal session, use kill_process instead.
Inputs: pid (required), device (required). Returns: data, text
get_configurationReMCP configurationARead-onlyIdempotentInspect
Use this when the person asks how the hosted ReMCP deployment/account is configured: server version and endpoint, runtime release installed on devices, retention/telemetry policy, hosted limits, and paired-computer versions. Use get_runtime_info for one computer's local runtime roots, policy and limits; use get_system_info for that computer's OS/CPU/memory/disk state.
Inputs: no named parameters. Returns: data, text
get_file_infoGet file infoARead-onlyIdempotentInspect
Use this when the user needs the size, type, permissions, or modification time of one file or directory before deciding what to do with it.
Inputs: path (required), device (required). Returns: data, text
get_more_search_resultsGet more search resultsARead-onlyIdempotentInspect
Use this when a search started with start_search still has more matches to return.
Inputs: device (required), sessionId (required), length, offset. Returns: data, text
get_recent_tool_callsRecent tool callsARead-onlyIdempotentInspect
Use this when you need to recover recent ReMCP activity or answer what was just done. It returns the most recent tool calls with the computer they ran on, success state and duration.
Inputs: tool, limit, device. Returns: data, text
get_runtime_infoGet runtime infoARead-onlyIdempotentInspect
Use this when you need one computer's local ReMCP runtime version, guardrails, allowed roots, or configured limits. Use get_system_info for host OS/CPU/memory/disk facts; deployment/account-level ReMCP server configuration belongs to get_configuration.
Inputs: device (required). Returns: data, text
get_runtime_statsGet runtime statsARead-onlyIdempotentInspect
Use this when you need counters from the current local runtime session on one machine: tool calls, failures, blocks, and active sessions. Account-level historical usage aggregated by device/tool over days belongs to get_usage_statistics.
Inputs: device (required). Returns: data, text
get_system_infoGet system infoARead-onlyIdempotentInspect
Use this when the user asks about machine health and host facts such as operating system, CPU load, memory, free disk space, or uptime. Use environment for PATH/shell/runtime environment details, and get_runtime_info for ReMCP-specific version, policy, and limits.
Inputs: device (required). Returns: data, text
get_usage_statisticsUsage statisticsARead-onlyIdempotentInspect
Use this when the person asks for account-level historical ReMCP usage over days: call counts, success rate, average duration, and per-tool/per-device breakdowns. Use get_runtime_stats instead for counters from the current local runtime session on one machine.
Inputs: days, tool, device. Returns: data, text
hash_fileHash fileARead-onlyIdempotentInspect
Use this when the user wants a checksum to verify a download, compare two files, or confirm a transfer.
Inputs: path (required), device (required), algorithm. Returns: data, text
installed_appsInstalled applicationsARead-onlyIdempotentInspect
Use this when the user needs to answer what software/packages are installed and their versions. Prefer a filter when looking for one product instead of returning the full inventory.
Inputs: device (required), limit, filter, refresh. Returns: data, text
interact_with_processInteract with processBDestructiveInspect
Use this when a running process is waiting for input, such as a prompt or a confirmation.
Inputs: pid (required), input (required), device (required), timeout_ms. Returns: data, text
keyboardKeyboard shortcutADestructiveInspect
Use this when the task calls for shortcuts and individual key presses such as Ctrl/Cmd+C, Tab, Enter or Escape. Do not use it to enter normal prose when type_text is available. On Wayland auto uses an already-authorized portal; backend=portal explicitly requests consent.
Inputs: device (required), app, key, pid, +5 more. Returns: data, text
kill_processKill processADestructiveInspect
Use this when the user wants to stop an arbitrary operating-system process by PID. If the PID belongs to a ReMCP terminal session created by start_process, prefer force_terminate so the session state and process tree are handled consistently.
Inputs: pid (required), device (required). Returns: data, text
launch_appLaunch applicationADestructiveInspect
Use this when the user explicitly asks to start a named desktop application or permitted executable on a paired computer. Launches the specified app with an argument array and no shell; start_process handles shell commands and scripts, while open_path opens existing permitted files or directories in their default application. Starting an app can execute code and change desktop state.
Inputs: device (required), app, cwd, args, +3 more. Returns: data, text
list_devicesList ReMCP devicesARead-onlyIdempotentInspect
Use this when the target computer is not already unambiguous. Call it first to list computers paired to the signed-in ReMCP account and see which ones are online.
Inputs: no named parameters. Returns: data, text
list_directoryList directoryARead-onlyIdempotentInspect
Use this when the user wants to see what a directory on a paired computer contains, optionally filtered by a glob.
Inputs: path (required), device (required), depth, pattern. Returns: data, text
list_processesList processesARead-onlyIdempotentInspect
Use this when the user wants the computer-wide operating-system process list. Use list_sessions when the question is only about commands started through ReMCP start_process.
Inputs: device (required), limit. Returns: data, text
list_searchesList searchesARead-onlyIdempotentInspect
Use this when you need to know which search sessions are still active on a machine.
Inputs: device (required). Returns: data, text
list_sessionsList sessionsARead-onlyIdempotentInspect
Use this when you need to know which ReMCP terminal sessions created by start_process are still running. Use list_processes for the computer-wide operating-system process list.
Inputs: device (required). Returns: data, text
list_windowsList windowsARead-onlyIdempotentInspect
Use this when the user needs to discover top-level desktop windows or obtain a stable window id/PID/bounds before window_action, screenshot_region or targeted input. Use ui_snapshot instead when you need controls inside a window; prefer list_windows over screenshot-based window guessing.
Inputs: device (required), include_monitor. Returns: data, text
move_fileMove or renameADestructiveInspect
Use this when the user asks to move or rename exactly one file or directory from one source to one destination. For two or more moves use move_paths.
Inputs: device (required), source (required), destination (required), overwrite. Returns: data, text
move_pathsMove pathsADestructiveInspect
Use this when the user asks to move or rename two or more files/directories in one batch. For one source/destination pair use move_file.
Inputs: paths (required), device (required), overwrite. Returns: data, text
move_to_trashMove to trashADestructiveInspect
Use this when the user wants a file or directory removed but still recoverable from the system trash.
Inputs: device (required), source (required). Returns: data, text
network_dnsNetwork DnsARead-onlyIdempotentInspect
Use this when the user asks to inspect the operating system configured DNS resolver server addresses on a paired computer. This reads local DNS configuration only; it performs no DNS lookup, external network request, or network configuration change.
Inputs: device (required). Returns: data, text
network_interfacesNetwork InterfacesARead-onlyIdempotentInspect
Use this when the user asks to inspect the full local network interface inventory on a paired computer, including OS-reported IPv4 and IPv6 addresses, netmasks, MAC addresses and internal-interface flags. Unlike network_summary, the inventory is not truncated. This only reads local adapter data; it does not scan networks or change network configuration.
Inputs: device (required). Returns: data, text
network_listenersNetwork ListenersARead-onlyIdempotentInspect
Use this when the user asks to inspect listening ports on a paired computer to diagnose local services, unexpected exposure or port conflicts. Reads the OS TCP listener inventory on Windows and macOS, and TCP/UDP listeners on Linux; process details depend on OS permissions. It does not connect to ports, scan remote hosts or change firewall rules.
Inputs: device (required). Returns: data, text
network_routesNetwork RoutesARead-onlyIdempotentInspect
Use this when the user asks to inspect the local OS routing table on a paired computer, including default gateway, destination networks, next hops and route metrics where supported. Use it to diagnose routing and reachability issues. The output varies by OS; it does not add routes, send traffic or change network settings.
Inputs: device (required). Returns: data, text
network_summaryNetwork SummaryARead-onlyIdempotentInspect
Use this when the user asks for a quick network overview of a paired computer: hostname, DNS servers and a bounded sample of its local network interfaces and IP addresses, including counts and truncation flags. Prefer network_interfaces for the complete inventory. This is read-only; it does not probe remote hosts or change network settings.
Inputs: device (required). Returns: data, text
network_testNetwork TestARead-onlyIdempotentInspect
Use this to check TCP connectivity from a paired computer to a specified host and port with a bounded timeout. It makes an outbound network connection but does not scan ranges or modify network settings.
Inputs: host (required), port (required), device (required), timeout_ms. Returns: data, text
notificationDesktop notificationAInspect
Use this when the user wants a visible native desktop notification on the paired computer. Do not use it as a substitute for replying in chat.
Inputs: device (required), message (required), title, timeout_seconds. Returns: data, text
open_pathOpen pathAInspect
Use this when the user wants the desktop operating system to open an existing permitted file or directory with its configured default application. The path is checked against the paired computer policy.
Inputs: path (required), device (required). Returns: data, text
pdf_annotationsPdf AnnotationsARead-onlyIdempotentInspect
Use this when the user asks to list or inspect annotations, comments, highlights or markup already present in one permitted local PDF file on a paired computer. Requires the PDF path and only returns annotation metadata; it does not edit, save, delete or annotate the PDF.
Inputs: path (required), device (required). Returns: data, text
pdf_extract_pagesPdf Extract PagesADestructiveInspect
Use this to extract a requested page range from a selected local PDF into an output PDF. Validate page numbers and output path; source pages remain unchanged.
Inputs: path (required), pages (required), device (required), output (required), +2 more. Returns: data, text
pdf_infoPdf InfoARead-onlyIdempotentInspect
Use this when the user asks to inspect an existing local PDF file on an explicitly paired computer and retrieve its metadata and properties, such as page count, title, author or PDF version when the local pdfinfo utility is available. Supply the permitted PDF path. Without pdfinfo, the fallback reports file size in bytes and annotation count. This is read-only: it does not extract page text, edit annotations, write output or replace the source PDF.
Inputs: path (required), device (required). Returns: data, text
pdf_mergePdf MergeADestructiveInspect
Use this to combine multiple user-selected local PDF files into a new output PDF in specified order. Check every input and destination; this creates an output file but does not edit the source pages.
Inputs: paths (required), device (required), output (required), path, +2 more. Returns: data, text
pdf_splitPdf SplitADestructiveInspect
Use this to split a selected local PDF into separate files or page groups using requested page boundaries. Verify the input and destination directory first; new output files are created.
Inputs: path (required), device (required), pages, output, +2 more. Returns: data, text
ping_devicePing a ReMCP deviceARead-onlyIdempotentInspect
Use this when you need to check that a paired computer answers right now and measure the round trip. The device agent answers the ping itself, so it works even while the local runtime is restarting or missing, and it reports hostname, platform, uptime and versions.
Inputs: device (required). Returns: data, text
pointer_clickPointer ClickADestructiveInspect
Use this to send one native mouse click at a verified desktop coordinate. Check the current screenshot or element bounds first; clicking can trigger purchases, deletion or other actions.
Inputs: device (required), x, y, button, +1 more. Returns: data, text
pointer_double_clickPointer Double ClickADestructiveInspect
Use this to send a native double-click at a verified desktop coordinate, for example to open a file or application. Inspect the target first; double-clicking may execute an action.
Inputs: device (required), x, y, button, +1 more. Returns: data, text
pointer_downPointer DownADestructiveInspect
Use this when the user asks to press and hold a mouse button to begin a native desktop drag or selection gesture on a paired computer. This sends a real pointer-button-down event to the current application and leaves the button held until pointer_up; it may change application state.
Inputs: device (required), x, y, button, +1 more. Returns: data, text
pointer_movePointer MoveAInspect
Use this to move the desktop mouse pointer to a requested coordinate without clicking. Prefer semantic UI tools for actions; movement alone does not select or submit anything.
Inputs: device (required), x, y, backend. Returns: data, text
pointer_right_clickPointer Right ClickADestructiveInspect
Use this to open a native context menu by right-clicking at a verified coordinate. Inspect the target and resulting menu before selecting any command.
Inputs: device (required), x, y, button, +1 more. Returns: data, text
pointer_upPointer UpADestructiveInspect
Use this when the user asks to release a mouse button after pointer_down during a native desktop drag or selection gesture on a paired computer. Releasing the button can complete a drag, click, selection or application action; it does not press the button.
Inputs: device (required), x, y, button, +1 more. Returns: data, text
power_lockPower LockAInspect
Use this to lock the paired computer screen while keeping the operating system and active applications running. Unlocking requires the device owner credentials.
Inputs: device (required), delay_seconds. Returns: data, text
power_restartPower RestartADestructiveInspect
Use this to reboot the paired computer when explicitly authorized. This ends active sessions and may interrupt unsaved work or remote access; verify necessity first.
Inputs: device (required), delay_seconds. Returns: data, text
power_shutdownPower ShutdownADestructiveInspect
Use this to shut down the paired computer when explicitly authorized. This ends running applications and remote connectivity; the device may require local power-on to reconnect.
Inputs: device (required), delay_seconds. Returns: data, text
power_sleepPower SleepADestructiveInspect
Use this to put the paired computer into system sleep mode when explicitly requested. Remote connectivity and active jobs may become unavailable until the device wakes.
Inputs: device (required), delay_seconds. Returns: data, text
read_binaryRead binary chunkARead-onlyIdempotentInspect
Use this when the user needs a non-text file transferred or inspected byte for byte as base64 chunks. For a normal image that should be shown or visually inspected, use read_image instead.
Inputs: path (required), device (required), length_bytes, offset_bytes. Returns: data, text
read_documentRead documentARead-onlyIdempotentInspect
Use this when the user needs structured document content from a PDF, DOCX, XLSX, CSV, JSON or XML file on a paired computer. Returns extracted content and bounded metadata from the document path.
Inputs: path (required), device (required), sheet, max_cells. Returns: data, text
read_fileRead fileARead-onlyIdempotentInspect
Use this when the task calls for plain text files on a paired computer: source code, configuration, Markdown, logs, scripts, or other line-oriented text. Use read_document instead for PDF, DOCX, XLSX and other structured document content. read_file keeps legacy PDF/DOCX extraction for compatibility but should not be the model’s first choice for those formats.
Inputs: path (required), device (required), length, offset. Returns: data, text
read_filesRead files by globARead-onlyIdempotentInspect
Use this when the exact file list is not already known and the user wants the contents of every file matching one glob under a directory. Use start_search instead to find filenames or text matches without reading every matched file. If exact paths are already known, use read_multiple_files; for one known path use read_file.
Inputs: device (required), path, pattern, max_files, +2 more. Returns: data, text
read_imageRead imageARead-onlyIdempotentInspect
Use this when the user wants to inspect or show an image that lives on a paired computer, such as a screenshot, photo, or diagram. It returns native MCP image content; do not use read_binary merely to display an image.
Inputs: path (required), device (required). Returns: data, text
read_multiple_filesRead multiple filesARead-onlyIdempotentInspect
Use this when the user names two or more exact file paths to read in one call. Do not use it for glob discovery: use read_files instead. For one known path use read_file.
Inputs: paths (required), device (required). Returns: data, text
read_process_outputRead process outputARead-onlyInspect
Use this when a process started earlier has more output to show and you do not want to wait for a pattern.
Inputs: pid (required), device (required), length, offset, +1 more. Returns: data, text
record_screenRecord screenADestructiveInspect
Use this when the task calls for a short bounded screen recording when motion/timing is necessary to diagnose a UI issue. Prefer screenshot_region for static visual verification because it is cheaper and easier to inspect.
Inputs: device (required), fps, destination, duration_seconds. Returns: data, text
remove_deviceRemove paired computerADestructiveInspect
Use this when the user needs to permanently remove an existing paired computer only when the user explicitly requests that irreversible action. Revokes the pairing; requires its device id.
Inputs: device (required). Returns: data, text
rename_deviceRename a computerAIdempotentInspect
Use this when the user wants to change the account-visible label of one paired computer. Do not use set_config_value for renaming: that device tool changes local runtime preferences instead. Access roots, command guardrails and resource limits stay on the computer itself.
Inputs: value (required), device (required), config_key (required). Returns: data, text
replace_in_filesReplace in filesADestructiveInspect
Use this when the same text or regular-expression replacement should be searched and applied across multiple files under a path, such as a rename or version bump. For one known block use edit_block; for one complete known file use write_file.
Inputs: path (required), device (required), pattern (required), replacement (required), +4 more. Returns: data, text
replace_linesReplace linesADestructiveInspect
Use this when the intended edit target is a specific 1-based line range in one file, optionally as a dry run. Use edit_block for matching an exact text block and write_file for replacing the complete file.
Inputs: path (required), device (required), content (required), end_line (required), +2 more. Returns: data, text
reveal_pathReveal pathAInspect
Use this when the user needs to reveal a permitted local file or directory in Finder/Explorer/file manager without opening the document itself.
Inputs: path (required), device (required). Returns: data, text
screenshot_regionScreenshot regionARead-onlyIdempotentInspect
Use this when static pixels are required for visual verification but only one window, monitor or rectangle matters. Prefer it over take_screenshot for a full desktop and over record_screen when motion/timing is not required; use ui_snapshot/browser_snapshot first when semantic structure is sufficient.
Inputs: device (required), x, y, app, +8 more. Returns: data, text
scrollScrollAInspect
Use this when the task calls for native wheel scrolling of a desktop region or semantic UI target. Prefer browser_action scroll_into_view for a known web element and ui_action scroll_into_view for a known accessible desktop element.
Inputs: device (required), x, y, id, +13 more. Returns: data, text
service_listService ListARead-onlyIdempotentInspect
Use this to list installed OS services on a paired computer for troubleshooting. Returns a local service inventory; it does not start, stop or modify services.
Inputs: device (required), name, scope. Returns: data, text
service_restartService RestartADestructiveInspect
Use this to restart one specifically named OS service on a paired computer when requested. This interrupts the service and can disrupt active users; check scope and recovery expectations first.
Inputs: name (required), device (required), scope. Returns: data, text
service_startService StartADestructiveInspect
Use this to start one specifically named OS service on a paired computer when requested. Starting may launch background processes and change system state; confirm name and scope first.
Inputs: name (required), device (required), scope. Returns: data, text
service_statusService StatusARead-onlyIdempotentInspect
Use this to inspect the current state of one named OS service on a paired computer. It reads status only and is preferable to service_list when the service name is known.
Inputs: name (required), device (required), scope. Returns: data, text
service_stopService StopADestructiveInspect
Use this to stop one specifically named OS service on a paired computer when requested. Stopping may interrupt users, applications or network availability; confirm the service name and impact first.
Inputs: name (required), device (required), scope. Returns: data, text
set_config_valueChange runtime settingAIdempotentInspect
Use this when the person wants to change a local runtime preference on one computer: telemetry opt-out, read/buffer line limits, or result size. Do not use this for account-level computer administration; ping_device, rename_device, enable_device, disable_device and remove_device cover account-level computer administration. Access roots, blocked commands, shell/guardrails and the write limit stay local-only.
Inputs: key (required), value (required), device (required). Returns: data, text
set_permissionsSet permissionsADestructiveInspect
Use this when the user asks to change mode bits or ownership of a file or directory, for example to make a script executable.
Inputs: mode (required), path (required), device (required), gid, +2 more. Returns: data, text
start_processStart processADestructiveInspect
Use this when the user asks to run a shell command, build, test suite, script, or terminal server that is not already covered by a narrower semantic tool. Prefer launch_app for GUI applications, service for service-manager operations, power_action for lock/sleep/restart/shutdown, network for routine network inspection/tests, event_log for OS logs, installed_apps for software inventory, environment for environment facts, and open_path/reveal_path for opening or revealing files.
Inputs: device (required), command (required), timeout_ms. Returns: data, text
start_searchStart searchAInspect
Use this when the user wants to find filenames or perform a content search under a path and does not need every matching file read in full. Use read_files instead when a glob is already known and the contents of every matched file are the desired result. Starts a search session and returns the first matches.
Inputs: path (required), device (required), pattern (required), ignoreCase, +7 more. Returns: data, text
stop_searchStop searchAIdempotentInspect
Use this when a search session is no longer needed and should release its resources.
Inputs: device (required), sessionId (required). Returns: data, text
take_screenshotTake screenshotAInspect
Use this when the user needs pixels for the entire desktop. Prefer ui_snapshot or browser_snapshot for semantic inspection, and screenshot_region for one window, monitor, or rectangle.
Inputs: device (required), keep, directory. Returns: data, text
type_textType textADestructiveInspect
Use this when the task calls for reliable Unicode text entry into a native semantic target or focused native control. Prefer method=auto so Accessibility is used when the target supports verified semantic editing. For Chromium page DOM, use browser_action instead of forcing desktop accessibility text input; auto fails safely when a semantic element rejects the write rather than claiming an unverified paste. Use method=keys only for deliberate physical key-by-key input, and keyboard for shortcuts/navigation/control keys.
Inputs: text (required), device (required), id, app, +14 more. Returns: data, text
ui_add_to_selectionUi Add To SelectionBDestructiveInspect
Use this to add an identified native UI item to the existing selection through Accessibility/UI Automation without clearing other selected items.
Inputs: device (required), id, app, pid, +5 more. Returns: data, text
ui_clickUi ClickADestructiveInspect
Use this to activate an identified native Accessibility/UI Automation element, such as a button or menu item, after ui_snapshot or ui_find. Clicking can submit data or execute destructive application actions.
Inputs: device (required), id, app, pid, +5 more. Returns: data, text
ui_collapseUi CollapseADestructiveInspect
Use this to collapse an identified native Accessibility/UI Automation tree node or disclosure control. It changes visible UI state without deleting child items.
Inputs: device (required), id, app, pid, +5 more. Returns: data, text
ui_expandUi ExpandADestructiveInspect
Use this to expand an identified native Accessibility/UI Automation tree node or disclosure control to reveal its children. Inspect the current element first.
Inputs: device (required), id, app, pid, +5 more. Returns: data, text
ui_findFind UI elementsARead-onlyIdempotentInspect
Use this when you need a small set of native UI elements by role/name/label/AutomationId after ui_snapshot. It requires an element criterion; use ui_snapshot, not an empty ui_find, to enumerate UI. Prefer returned id or compact label for subsequent ui_action; refresh only when the UI has changed.
Inputs: device (required), id, app, pid, +9 more. Returns: data, text
ui_focusUi FocusAInspect
Use this to focus an identified native Accessibility/UI Automation field or control prior to keyboard input. It does not insert text; identify the element from a recent snapshot.
Inputs: device (required), id, app, pid, +5 more. Returns: data, text
ui_invokeUi InvokeADestructiveInspect
Use this when the user asks to invoke one identified native Accessibility/UI Automation element, such as a button or menu command, by its element ID or a specific role/name selector on a paired computer. Find the control with ui_snapshot or ui_find first. Invoking a control can submit data or cause destructive application actions.
Inputs: device (required), id, app, pid, +5 more. Returns: data, text
ui_remove_from_selectionUi Remove From SelectionADestructiveInspect
Use this to deselect one identified native UI item through Accessibility/UI Automation without removing the underlying data or clearing all selections.
Inputs: device (required), id, app, pid, +5 more. Returns: data, text
ui_scroll_into_viewUi Scroll Into ViewAInspect
Use this to bring an identified native Accessibility/UI Automation element into view before interacting with it. Prefer semantic targeting over blind pointer scrolling.
Inputs: device (required), id, app, pid, +5 more. Returns: data, text
ui_selectUi SelectADestructiveInspect
Use this to select an option in an identified native Accessibility/UI Automation list or selection control. Verify available options first; selection may change application state.
Inputs: device (required), id, app, pid, +6 more. Returns: data, text
ui_set_range_valueUi Set Range ValueADestructiveInspect
Use this to assign a supported value to an identified native slider or range widget through Accessibility/UI Automation. Inspect allowed bounds first; changes can take effect immediately.
Inputs: device (required), id, app, pid, +6 more. Returns: data, text
ui_set_valueUi Set ValueADestructiveInspect
Use this to replace the value of a native accessible text field or editable control by element ID or semantic selector. Check the target before writing; it may update application data.
Inputs: device (required), id, app, pid, +6 more. Returns: data, text
ui_snapshotUI accessibility snapshotARead-onlyIdempotentInspect
Use this when the user needs to inspect native desktop Accessibility/UI Automation semantics for the active app before clicking or typing. Prefer scope=active; use scope=desktop only for cross-application discovery. For a Chromium web page with CDP available, prefer browser_snapshot for page DOM semantics.
Inputs: device (required), app, pid, scope, +5 more. Returns: data, text
ui_toggleUi ToggleADestructiveInspect
Use this to activate an identified native checkbox or switch using its accessibility semantics. Check the present state before toggling; this can change saved settings.
Inputs: device (required), id, app, pid, +5 more. Returns: data, text
wait_for_process_outputWait for process outputARead-onlyInspect
Use this when you need to wait until a running command prints something specific, such as a readiness line or a failing test name.
Inputs: pid (required), device (required), pattern (required), timeout_ms. Returns: data, text
wait_for_uiWait for UIARead-onlyIdempotentInspect
Use this when a desktop action must be followed by proof of an explicit native UI state or condition instead of sleeping or polling screenshots. Presence/absence waits need a semantic target; use state=changed for a broad tree change and set a bounded timeout.
Inputs: device (required), id, app, pid, +11 more. Returns: data, text
who_am_iWho am IARead-onlyIdempotentInspect
Use this when the person asks which ReMCP account they are signed in as, how many machines are paired and online, or how many tool calls they have made lately.
Inputs: no named parameters. Returns: data, text
window_closeWindow CloseADestructiveInspect
Use this to close a specific application window on the paired computer. Closing may discard unsaved changes or quit an application; confirm the intended window and save work first.
Inputs: device (required), id, app, pid, +2 more. Returns: data, text
window_focusWindow FocusAInspect
Use this to activate an existing application window on the paired desktop, identified by a current window ID, process or title. This only changes the active window; use list_windows if the target is unknown.
Inputs: device (required), id, app, pid, +2 more. Returns: data, text
window_maximizeWindow MaximizeAInspect
Use this to maximize an identified application window on the paired desktop. Confirm the target window first; it changes window layout without modifying application files.
Inputs: device (required), id, app, pid, +2 more. Returns: data, text
window_minimizeWindow MinimizeAInspect
Use this to minimize a selected application window while leaving the application running. Identify the window first; it does not close the window or terminate its process.
Inputs: device (required), id, app, pid, +2 more. Returns: data, text
window_moveWindow MoveAInspect
Use this to reposition an identified desktop window to specified screen coordinates. Confirm monitor geometry before moving; the application continues running.
Inputs: x (required), y (required), device (required), id, +4 more. Returns: data, text
window_move_resizeWindow Move ResizeBInspect
Use this to reposition and resize a specific desktop window in one action with explicit coordinates and dimensions. Verify its identity and monitor bounds beforehand.
Inputs: x (required), y (required), width (required), device (required), +6 more. Returns: data, text
window_resizeWindow ResizeAInspect
Use this to resize an identified desktop window to the requested width and height. Inspect display limits first; it does not change the content of files in the application.
Inputs: width (required), device (required), height (required), id, +4 more. Returns: data, text
window_restoreWindow RestoreAInspect
Use this to restore an identified minimized or maximized application window to its normal state. It changes window state but does not restore deleted data.
Inputs: device (required), id, app, pid, +2 more. Returns: data, text
write_binaryWrite binary chunkADestructiveInspect
Use this when the user needs base64 data written to a file on a paired computer byte for byte, including files too large for a single call.
Inputs: data (required), path (required), device (required), mode. Returns: data, text
write_fileWrite fileADestructiveInspect
Use this when the user asks to create or replace the complete contents of exactly one text file. For a small exact-block edit use edit_block, for a known line range use replace_lines, and when you already have a unified diff use apply_patch. For two or more independent complete files use write_files; use write_binary for binary bytes.
Inputs: path (required), device (required), content (required), mode. Returns: data, text
write_filesWrite multiple filesADestructiveInspect
Use this when the user asks to create or replace the complete contents of two or more independent text files in one batch. For one path use write_file; for partial edits use edit_block, replace_lines, replace_in_files, or apply_patch instead of rewriting whole files.
Inputs: files (required), device (required). Returns: data, text
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Secure AI access to Windows, macOS, Linux computers and a hosted Cloud Browser.
Use a person's computer after they approve it there: files, Terminal, browser, local MCP.
Use a person's computer after they approve it there: files, Terminal, browser, local MCP.
Give your AI hands on every computer you own. Remote control of your Windows, Mac and Linux machines from one ChatGPT, Claude or Cursor chat — even from your phone. Ask it to find out why the office server is crawling, update every PC in the house, sort out your mom's laptop from another city, or ship the site from your work machine while you're on a train. It runs commands, reads files and takes a screenshot when it needs to see the screen. You decide what each computer allows, and you can cut any of them off in a second.
Related MCP Servers
- AlicenseNot gradedqualityBmaintenanceEnables AI assistants like Claude, Codex and ChatGPT to actually operate a computer — finding files, running commands, taking screenshots and driving a browser — and, after a one-time pairing, to help on family members' computers anywhere. Connections between machines are end-to-end encrypted through a relay, requiring no VPN, port forwarding or router setup.1MIT
- AlicenseNot gradedqualityCmaintenanceEnables AI clients to securely control trusted devices and perform filesystem, terminal, process, and document operations through a self-hosted remote service.1,287,096 npmMIT
- AlicenseNot gradedqualityBmaintenancePre-alpha P2P remote-control app with a local stdio MCP bridge for paired computers. Users build from source; platform support and host-side approvals apply.21AGPL 3.0
- AlicenseNot gradedqualityBmaintenanceIt enables secure local monitoring and control of Windows PCs, including CPU/RAM/GPU metrics, process management, screenshots, clipboard, volume, power actions, and sandboxed file listing within allowed directories.MIT
Glama MCP Gateway
Add one secure layer between your agents and this server.