@github/computer-use-mcp 0.1.9 → 0.1.12

package/dist/index.d.ts

@@ -3,8 +3,8 @@ import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 /**
  * Creates a computer-use MCP server.
  *
- * The server exposes a single "computer" tool with mouse, keyboard, and
- * screenshot actions mirroring the Anthropic computer-use API.
+ * The server exposes individual tools for mouse, keyboard, and screenshot
+ * actions (screenshot, left_click, type, key, scroll, etc.).
  *
  * Uses prebuilt native addons (macOS / Windows / Linux). Throws when the
  * current platform or desktop session is unsupported.

package/dist/index.js

@@ -1,20 +1,26 @@
-import{McpServer as N}from"@modelcontextprotocol/sdk/server/mcp.js";import{createRequire as k}from"module";import{dirname as w,join as P}from"path";import{fileURLToPath as $}from"url";var C=k(import.meta.url),R=w($(import.meta.url));function q(){let c=P(R,"prebuilds",`${process.platform}-${process.arch}`,"computer.node");try{return C(c)}catch{return null}}var b=class c{constructor(e){this.native=e}native;static create(){let e=q();if(!e)throw new Error(`Native computer bindings not available for ${process.platform}-${process.arch}`);return e.display(),new c(e)}async click(e,o,s,t){this.native.click(e,o,s,t)}async move(e,o){this.native.move(e,o)}async drag(e,o,s,t){this.native.drag(e,o,s,t)}async mouseDown(e,o){this.native.mouseDown(e,o)}async mouseUp(e,o){this.native.mouseUp(e,o)}async type(e){this.native.type(e)}async key(e){this.native.key(e)}async scroll(e,o,s,t){this.native.scroll(e,o,s,t)}async cursorPosition(){return this.native.cursorPosition()}async display(){let{width:e,height:o}=this.native.display();return{width:e,height:o}}async screenshot(){return this.native.screenshot()}};import{z as a}from"zod";function i(c,e){let o=[{type:"text",text:c}];return e?{content:o,isError:!0}:{content:o}}function T(c,e){return{content:[{type:"text",text:c},{type:"image",data:e,mimeType:"image/png"}]}}var S=a.enum(["screenshot","left_click","right_click","middle_click","double_click","triple_click","left_click_drag","mouse_move","type","key","scroll","left_mouse_down","left_mouse_up","hold_key","wait","cursor_position"]),x=a.object({x:a.number(),y:a.number()});function h(c,e){c.tool("computer",`Use a mouse and keyboard to interact with a computer, and take screenshots.
-
-Take a screenshot before clicking to determine coordinates. After each action, take a screenshot to verify the result.`,{action:S.describe(`The action to perform:
-- screenshot: Capture the current screen. No other parameters.
-- left_click: Click left button. Optional 'coordinate' (clicks at current position if omitted).
-- right_click: Click right button. Optional 'coordinate'.
-- middle_click: Click middle button. Optional 'coordinate'.
-- double_click: Double-click left button. Optional 'coordinate'.
-- triple_click: Triple-click left button. Optional 'coordinate'.
-- left_click_drag: Drag from 'start_coordinate' to 'coordinate'. Both required.
-- mouse_move: Move cursor. Requires 'coordinate'.
-- type: Type text. Requires 'text'. Rejects 'coordinate'.
-- key: Press key combo. Requires 'text' (e.g. 'Return', 'ctrl+s', 'alt+Tab'). Rejects 'coordinate'.
-- scroll: Scroll. Requires 'scroll_direction'. Optional 'coordinate', 'scroll_amount' (default 3).
-- left_mouse_down: Press left button at current position. No 'coordinate'.
-- left_mouse_up: Release left button at current position. No 'coordinate'.
-- hold_key: Hold a key. Requires 'text' (key name) and 'duration' (seconds, max 100).
-- wait: Pause. Requires 'duration' (seconds, max 100).
-- cursor_position: Get current cursor {x, y}. No other parameters.`),coordinate:x.optional().describe("{x, y} pixel coordinate. Required for mouse_move and left_click_drag (end position). Optional for clicks and scroll."),start_coordinate:x.optional().describe("{x, y} drag start position. Required for left_click_drag only."),text:a.string().optional().describe("For 'type': text to type. For 'key': key combo (e.g. 'ctrl+s'). For 'hold_key': key name."),scroll_direction:a.enum(["up","down","left","right"]).optional().describe("Required for 'scroll' action."),scroll_amount:a.number().int().nonnegative().optional().describe("Scroll clicks (default 3). Only for 'scroll'."),duration:a.number().nonnegative().max(100).optional().describe("Seconds. Required for 'hold_key' and 'wait'.")},async o=>{let{action:s,coordinate:t,start_coordinate:d,text:u,scroll_direction:m,scroll_amount:v,duration:p}=o,n=await e();switch(s){case"screenshot":{let r=await n.screenshot();if(!r)return i("Screenshot failed",!0);let l=await n.display();return T(`Display: ${l.width}x${l.height}`,r.toString("base64"))}case"cursor_position":{let r=await n.cursorPosition();return i(`${r.x},${r.y}`)}case"left_click":case"right_click":case"middle_click":case"double_click":case"triple_click":{let r=s.includes("right")?"right":s.includes("middle")?"middle":"left",l=s.includes("double")?2:s.includes("triple")?3:1;t&&await n.move(t.x,t.y);let y=t??await n.cursorPosition();return await n.click(y.x,y.y,r,l),i(`${s}${t?` at (${t.x},${t.y})`:""}`)}case"left_click_drag":return!d||!t?i("left_click_drag requires start_coordinate and coordinate",!0):(await n.drag(d.x,d.y,t.x,t.y),i(`Dragged (${d.x},${d.y}) -> (${t.x},${t.y})`));case"mouse_move":return t?(await n.move(t.x,t.y),i(`Moved to (${t.x},${t.y})`)):i("mouse_move requires coordinate",!0);case"type":return u?(await n.type(u),i(`Typed ${u.length} chars`)):i("type requires text",!0);case"key":return u?(await n.key(u),i(`Pressed ${u}`)):i("key requires text",!0);case"scroll":{if(!m)return i("scroll requires scroll_direction",!0);let r=v??3,l=await n.cursorPosition(),y=t?.x??l.x,f=t?.y??l.y,g=m==="left"?-r:m==="right"?r:0,_=m==="down"?r:m==="up"?-r:0;return await n.scroll(y,f,g,_),i(`Scrolled ${m} ${r} at (${y},${f})`)}case"left_mouse_down":{let r=await n.cursorPosition();return await n.mouseDown(r.x,r.y),i(`Mouse down at (${r.x},${r.y})`)}case"left_mouse_up":{let r=await n.cursorPosition();return await n.mouseUp(r.x,r.y),i(`Mouse up at (${r.x},${r.y})`)}case"hold_key":return!u||p==null?i("hold_key requires text and duration",!0):(await n.key(u),i(`Held ${u} for ${p}s`));case"wait":return p==null?i("wait requires duration",!0):(await new Promise(r=>setTimeout(r,p*1e3)),i(`Waited ${p}s`));default:return i(`Unknown action: ${String(s)}`,!0)}})}function E(){let c=new N({name:"computer-use",version:"1.0.0"}),e;return h(c,async()=>(e||(e=b.create()),e)),c}export{E as createServer};
+import{McpServer as A}from"@modelcontextprotocol/sdk/server/mcp.js";import{existsSync as P,mkdirSync as k,readFileSync as D,writeFileSync as y}from"fs";import{homedir as T}from"os";import{join as g}from"path";var v="<!-- LOG -->",h=class{constructor(e,t=g(T(),".cache","github","computer-use-mcp","logs")){this.inner=e;this.logDir=t,this.path=g(this.logDir,`${this.timestamp()}.html`),P(this.logDir)||k(this.logDir,{recursive:!0}),y(this.path,`<!DOCTYPE html>
+<html>
+<head>
+<meta charset="utf-8">
+<style>
+  body { font-family: system-ui, sans-serif; margin: 20px; background: #fafafa; }
+  table { border-collapse: collapse; width: 100%; }
+  th, td { border: 1px solid #ddd; padding: 6px 10px; text-align: left; vertical-align: top; }
+  th { background: #f0f0f0; }
+  img { max-width: 800px; cursor: pointer; }
+  .time { color: #888; white-space: nowrap; }
+  .action { font-weight: 600; }
+  .error { color: #c00; }
+</style>
+</head>
+<body>
+<table>
+<tr><th>Time</th><th>Action</th><th>Arguments</th></tr>
+${v}
+</table>
+</body>
+</html>
+`)}inner;logDir;path;startTime=performance.now();timestamp(){let e=new Date,t=(o,i=2)=>String(o).padStart(i,"0");return`${e.getFullYear()}${t(e.getMonth()+1)}${t(e.getDate())}-${t(e.getHours())}${t(e.getMinutes())}${t(e.getSeconds())}.${t(Math.floor(e.getMilliseconds()/10))}`}elapsed(){return`${((performance.now()-this.startTime)/1e3).toFixed(2)}s`}insert(e){let t=D(this.path,"utf-8");y(this.path,t.replace(v,e+v))}row(e,t){this.insert(`<tr><td class="time">${this.elapsed()}</td><td class="action">${e}</td><td>${t}</td></tr>
+`)}async click(e,t,o,i){await this.inner.click(e,t,o,i),this.row("click",`(${e}, ${t}) button=${o} count=${i}`)}async move(e,t){await this.inner.move(e,t),this.row("move",`(${e}, ${t})`)}async drag(e,t,o,i){await this.inner.drag(e,t,o,i),this.row("drag",`(${e}, ${t}) \u2192 (${o}, ${i})`)}async mouseDown(e,t){await this.inner.mouseDown(e,t),this.row("mouseDown",`(${e}, ${t})`)}async mouseUp(e,t){await this.inner.mouseUp(e,t),this.row("mouseUp",`(${e}, ${t})`)}async type(e){await this.inner.type(e);let t=e.replace(/&/g,"&amp;").replace(/</g,"&lt;").replace(/>/g,"&gt;").replace(/"/g,"&quot;");this.row("type",`"${t}"`)}async key(e){await this.inner.key(e),this.row("key",e)}async scroll(e,t,o,i){await this.inner.scroll(e,t,o,i),this.row("scroll",`(${e}, ${t}) dx=${o} dy=${i}`)}async cursorPosition(){let e=await this.inner.cursorPosition();return this.row("cursorPosition",`(${e.x}, ${e.y})`),e}async display(){let e=await this.inner.display();return this.row("display",`${e.width}\xD7${e.height}`),e}async screenshot(e,t){let o=await this.inner.screenshot(e,t);if(o){let i=`${this.timestamp()}.png`;y(g(this.logDir,i),o);let r=e&&t?` (${e}\xD7${t})`:"";this.insert(`<tr><td class="time">${this.elapsed()}</td><td class="action">screenshot</td><td>${r}<br><img src="${i}"></td></tr>
+`)}else this.row("screenshot",'<span class="error">failed</span>');return o}};import{createRequire as N}from"module";import{dirname as S,join as C}from"path";import{fileURLToPath as W}from"url";var M=N(import.meta.url),I=S(W(import.meta.url)),b=class n{constructor(e){this.native=e}native;scaleX=1;scaleY=1;static create(){let e=C(I,"prebuilds",`${process.platform}-${process.arch}`,"computer.node"),t;try{t=M(e)}catch{throw new Error(`Native computer bindings not available for ${process.platform}-${process.arch}`)}return t.display(),new n(t)}toNative(e,t){return[e/this.scaleX,t/this.scaleY]}fromNative(e,t){return[Math.round(e*this.scaleX),Math.round(t*this.scaleY)]}async click(e,t,o,i){let[r,s]=this.toNative(e,t);this.native.click(r,s,o,i)}async move(e,t){let[o,i]=this.toNative(e,t);this.native.move(o,i)}async drag(e,t,o,i){let[r,s]=this.toNative(e,t),[u,a]=this.toNative(o,i);this.native.drag(r,s,u,a)}async mouseDown(e,t){let[o,i]=this.toNative(e,t);this.native.mouseDown(o,i)}async mouseUp(e,t){let[o,i]=this.toNative(e,t);this.native.mouseUp(o,i)}async type(e){this.native.type(e)}async key(e){this.native.key(e)}async scroll(e,t,o,i){let[r,s]=this.toNative(e,t);this.native.scroll(r,s,o,i)}async cursorPosition(){let e=this.native.cursorPosition(),[t,o]=this.fromNative(e.x,e.y);return{x:t,y:o}}async display(){let{width:e,height:t}=this.native.display();return{width:e,height:t}}async screenshot(e,t){let{width:o,height:i}=this.native.display(),r,s;if(e&&e>0&&t&&t>0)r=e,s=t;else{let l=Math.min(1,1568/Math.max(o,i),Math.sqrt(115e4/(o*i)));r=Math.floor(o*l),s=Math.floor(i*l)}return this.scaleX=r/o,this.scaleY=s/i,r===o&&s===i?this.native.screenshot(0,0):this.native.screenshot(r,s)}async capabilities(){return this.native.capabilities()}async listDisplays(){return this.native.listDisplays()}async captureDisplay(e,t=[]){return this.native.captureDisplay(e,t)}async listWindows(){return this.native.listWindows()}async activateWindow(e){return this.native.activateWindow(e)}async minimizeWindow(e){return this.native.minimizeWindow(e)}async restoreWindow(e){return this.native.restoreWindow(e)}};import{z as m}from"zod";function c(n,e){let t=[{type:"text",text:n}];return e?{content:t,isError:!0}:{content:t}}var d=m.object({x:m.number().describe("Horizontal pixel coordinate."),y:m.number().describe("Vertical pixel coordinate.")});function f(n,e){n.registerTool("screenshot",{description:"Capture the current screen. Take a screenshot before clicking to determine coordinates and after each action to verify the result.",inputSchema:{width:m.number().int().positive().optional().describe("Optional target width in pixels for resizing the screenshot."),height:m.number().int().positive().optional().describe("Optional target height in pixels for resizing the screenshot.")}},async({width:i,height:r})=>{let u=await(await e()).screenshot(i,r);return u?{content:[{type:"image",data:u.toString("base64"),mimeType:"image/png"}]}:c("Screenshot failed",!0)}),n.registerTool("cursor_position",{description:"Get the current cursor position in pixel coordinates. Returns {x, y}."},async()=>{let r=await(await e()).cursorPosition();return c(`${r.x},${r.y}`)});let t={coordinate:d.optional().describe("{x, y} pixel coordinate to click at. Clicks at current cursor position if omitted.")},o=(i,r,s,u)=>{n.registerTool(i,{description:r,inputSchema:t},async({coordinate:a})=>{let l=await e();a&&await l.move(a.x,a.y);let p=a??await l.cursorPosition();return await l.click(p.x,p.y,s,u),c(`${i}${a?` at (${a.x},${a.y})`:""}`)})};o("left_click","Click the left mouse button. Optionally move to a coordinate first.","left",1),o("right_click","Click the right mouse button. Optionally move to a coordinate first.","right",1),o("middle_click","Click the middle mouse button. Optionally move to a coordinate first.","middle",1),o("double_click","Double-click the left mouse button. Optionally move to a coordinate first.","left",2),o("triple_click","Triple-click the left mouse button. Optionally move to a coordinate first.","left",3),n.registerTool("mouse_move",{description:"Move the mouse cursor to the specified pixel coordinate.",inputSchema:{coordinate:d.describe("{x, y} pixel coordinate to move the cursor to.")}},async({coordinate:i})=>(await(await e()).move(i.x,i.y),c(`Moved to (${i.x},${i.y})`))),n.registerTool("left_click_drag",{description:"Click and drag from a start coordinate to an end coordinate.",inputSchema:{start_coordinate:d.describe("{x, y} pixel coordinate to start the drag from."),coordinate:d.describe("{x, y} pixel coordinate to drag to.")}},async({start_coordinate:i,coordinate:r})=>(await(await e()).drag(i.x,i.y,r.x,r.y),c(`Dragged (${i.x},${i.y}) -> (${r.x},${r.y})`))),n.registerTool("left_mouse_down",{description:"Press and hold the left mouse button at the current cursor position."},async()=>{let i=await e(),r=await i.cursorPosition();return await i.mouseDown(r.x,r.y),c(`Mouse down at (${r.x},${r.y})`)}),n.registerTool("left_mouse_up",{description:"Release the left mouse button at the current cursor position."},async()=>{let i=await e(),r=await i.cursorPosition();return await i.mouseUp(r.x,r.y),c(`Mouse up at (${r.x},${r.y})`)}),n.registerTool("type",{description:"Type a string of text using the keyboard.",inputSchema:{text:m.string().describe("The text to type.")}},async({text:i})=>(await(await e()).type(i),c(`Typed ${i.length} chars`))),n.registerTool("key",{description:"Press a key or key combination. Examples: 'Return', 'ctrl+s', 'alt+Tab', 'BackSpace'.",inputSchema:{text:m.string().describe("Key combo to press (e.g. 'Return', 'ctrl+s', 'alt+Tab').")}},async({text:i})=>(await(await e()).key(i),c(`Pressed ${i}`))),n.registerTool("hold_key",{description:"Hold a key down for a specified duration in seconds.",inputSchema:{text:m.string().describe("Key name to hold (e.g. 'shift', 'ctrl')."),duration:m.number().nonnegative().max(100).describe("Seconds to hold the key (max 100).")}},async({text:i,duration:r})=>(await(await e()).key(i),c(`Held ${i} for ${r}s`))),n.registerTool("scroll",{description:"Scroll the screen in a given direction at an optional coordinate.",inputSchema:{scroll_direction:m.enum(["up","down","left","right"]).describe("Direction to scroll."),coordinate:d.optional().describe("{x, y} pixel coordinate to scroll at. Scrolls at current cursor position if omitted."),scroll_amount:m.number().int().nonnegative().optional().describe("Number of scroll clicks (default 3).")}},async({scroll_direction:i,coordinate:r,scroll_amount:s})=>{let u=await e(),a=s??3,l=await u.cursorPosition(),p=r?.x??l.x,w=r?.y??l.y,x=i==="left"?-a:i==="right"?a:0,$=i==="down"?a:i==="up"?-a:0;return await u.scroll(p,w,x,$),c(`Scrolled ${i} ${a} at (${p},${w})`)}),n.registerTool("wait",{description:"Pause for a specified duration in seconds.",inputSchema:{duration:m.number().nonnegative().max(100).describe("Seconds to wait (max 100).")}},async({duration:i})=>(await new Promise(r=>setTimeout(r,i*1e3)),c(`Waited ${i}s`)))}function G(){let n=new A({name:"computer-use",version:"1.0.0"}),e=!1,t;return f(n,async()=>(t||(t=b.create(),e&&(t=new h(t))),t)),n}export{G as createServer};
 //# sourceMappingURL=index.js.map

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@github/computer-use-mcp",
-  "version": "0.1.9",
+  "version": "0.1.12",
   "description": "Computer Use MCP Server",
   "author": "GitHub",
   "repository": {