@github/computer-use-mcp 0.1.9 → 0.1.10

package/dist/index.js

@@ -1,6 +1,6 @@
-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.
+import{McpServer as S}from"@modelcontextprotocol/sdk/server/mcp.js";import{createRequire as k}from"module";import{dirname as w,join as P}from"path";import{fileURLToPath as $}from"url";var C=k(import.meta.url),q=w($(import.meta.url));function R(){let c=P(q,"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=R();if(!e)throw new Error(`Native computer bindings not available for ${process.platform}-${process.arch}`);return e.display(),new c(e)}async click(e,i,s,t){this.native.click(e,i,s,t)}async move(e,i){this.native.move(e,i)}async drag(e,i,s,t){this.native.drag(e,i,s,t)}async mouseDown(e,i){this.native.mouseDown(e,i)}async mouseUp(e,i){this.native.mouseUp(e,i)}async type(e){this.native.type(e)}async key(e){this.native.key(e)}async scroll(e,i,s,t){this.native.scroll(e,i,s,t)}async cursorPosition(){return this.native.cursorPosition()}async display(){let{width:e,height:i}=this.native.display();return{width:e,height:i}}async screenshot(){return this.native.screenshot()}};import{z as a}from"zod";function o(c,e){let i=[{type:"text",text:c}];return e?{content:i,isError:!0}:{content:i}}var T=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"]),v=a.object({x:a.number(),y:a.number()});function x(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:
+Take a screenshot before clicking to determine coordinates. After each action, take a screenshot to verify the result.`,{action:T.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'.
@@ -16,5 +16,5 @@ Take a screenshot before clicking to determine coordinates. After each action, t
 - 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: ${String(s)}`,!0)}})}function E(){let c=new N({name:"computer-use",version:"1.0.0"}),e;return h(c,async()=>(e||(e=b.create()),e)),c}export{E as createServer};
+- cursor_position: Get current cursor {x, y}. No other parameters.`),coordinate:v.optional().describe("{x, y} pixel coordinate. Required for mouse_move and left_click_drag (end position). Optional for clicks and scroll."),start_coordinate:v.optional().describe("{x, y} drag start position. Required for left_click_drag only."),text:a.string().optional().describe("For 'type': text to type. For 'key': key combo (e.g. 'ctrl+s'). For 'hold_key': key name."),scroll_direction:a.enum(["up","down","left","right"]).optional().describe("Required for 'scroll' action."),scroll_amount:a.number().int().nonnegative().optional().describe("Scroll clicks (default 3). Only for 'scroll'."),duration:a.number().nonnegative().max(100).optional().describe("Seconds. Required for 'hold_key' and 'wait'.")},async i=>{let{action:s,coordinate:t,start_coordinate:d,text:u,scroll_direction:m,scroll_amount:h,duration:p}=i,n=await e();switch(s){case"screenshot":{let r=await n.screenshot();return r?{content:[{type:"image",data:r.toString("base64"),mimeType:"image/png"}]}:o("Screenshot failed",!0)}case"cursor_position":{let r=await n.cursorPosition();return o(`${r.x},${r.y}`)}case"left_click":case"right_click":case"middle_click":case"double_click":case"triple_click":{let r=s.includes("right")?"right":s.includes("middle")?"middle":"left",y=s.includes("double")?2:s.includes("triple")?3:1;t&&await n.move(t.x,t.y);let l=t??await n.cursorPosition();return await n.click(l.x,l.y,r,y),o(`${s}${t?` at (${t.x},${t.y})`:""}`)}case"left_click_drag":return!d||!t?o("left_click_drag requires start_coordinate and coordinate",!0):(await n.drag(d.x,d.y,t.x,t.y),o(`Dragged (${d.x},${d.y}) -> (${t.x},${t.y})`));case"mouse_move":return t?(await n.move(t.x,t.y),o(`Moved to (${t.x},${t.y})`)):o("mouse_move requires coordinate",!0);case"type":return u?(await n.type(u),o(`Typed ${u.length} chars`)):o("type requires text",!0);case"key":return u?(await n.key(u),o(`Pressed ${u}`)):o("key requires text",!0);case"scroll":{if(!m)return o("scroll requires scroll_direction",!0);let r=h??3,y=await n.cursorPosition(),l=t?.x??y.x,f=t?.y??y.y,g=m==="left"?-r:m==="right"?r:0,_=m==="down"?r:m==="up"?-r:0;return await n.scroll(l,f,g,_),o(`Scrolled ${m} ${r} at (${l},${f})`)}case"left_mouse_down":{let r=await n.cursorPosition();return await n.mouseDown(r.x,r.y),o(`Mouse down at (${r.x},${r.y})`)}case"left_mouse_up":{let r=await n.cursorPosition();return await n.mouseUp(r.x,r.y),o(`Mouse up at (${r.x},${r.y})`)}case"hold_key":return!u||p==null?o("hold_key requires text and duration",!0):(await n.key(u),o(`Held ${u} for ${p}s`));case"wait":return p==null?o("wait requires duration",!0):(await new Promise(r=>setTimeout(r,p*1e3)),o(`Waited ${p}s`));default:return o(`Unknown action: ${String(s)}`,!0)}})}function H(){let c=new S({name:"computer-use",version:"1.0.0"}),e;return x(c,async()=>(e||(e=b.create()),e)),c}export{H as createServer};
 //# sourceMappingURL=index.js.map

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@github/computer-use-mcp",
-  "version": "0.1.9",
+  "version": "0.1.10",
   "description": "Computer Use MCP Server",
   "author": "GitHub",
   "repository": {