Windows 11 Error: Worker service failed to start on port 37777. (port 37777)

[AnemoneIndicum]

⚡ Quick Bug Report (Recommended)

I updated the plugin to version 7.3.9 today, but when I started Claude Code for the first time after the update, I encountered the error: "Error: Worker service failed to start on port 37777. (port 37777)". I've mentioned this issue in previous issues as well. Previously, the service corresponding to port 37777 would run after Claude Code started, although it would be inaccessible intermittently during Claude programming. Unfortunately, after today's update, it won't start, even after using npm run worker:restart and npm run worker:start. I hope the author can fix this issue soon, resolving this annoying problem for such a great project.
#363
#361

Environment

• Claude-mem version: 7.3.9
• Claude Code version: 2.0.61
• OS: Windows
• Platform:

Logs

Worker logs are located at:

• Path: ~/.claude-mem/logs/worker-YYYY-MM-DD.log
• Example: ~/.claude-mem/logs/worker-2025-12-14.log

Please paste relevant log entries (last 50 lines or error messages):

  /model to try Opus 4.5
  ⎿  SessionStart:Callback hook succeeded: Success
  ⎿  SessionStart:startup says: Plugin hook "node
     "C:\Users\BAIJU\.claude\plugins\marketplaces\thedotmack\plugin/scripts/user-message-hook.js"" failed to start: The
     operation was aborted. Check that the command exists and is executable.
  ⎿  SessionStart:Callback hook succeeded: Success
  ⎿ SessionStart:startup says: Plugin hook error: [2025-12-18 11:36:12.677] [ERROR] [SYSTEM] Failed to start worker
    {platform=win32, port=37777, error=Worker failed to start on Windows (readiness check timed out after 20000ms)

    Troubleshooting:
    1. Check Task Manager for zombie 'bun.exe' or 'node.exe' processes
    2. Verify port 37777 is not in use: netstat -ano | findstr 37777
    3. Check worker logs in ~/.claude-mem/logs/
    4. See GitHub issues: #363, #367, #371, #373
    5. Docs: https://docs.claude-mem.ai/troubleshooting/windows-issues,
    marketplaceRoot=C:\Users\BAIJU\.claude\plugins\marketplaces\thedotmack}
    file:///C:/Users/BAIJU/.claude/plugins/marketplaces/thedotmack/plugin/scripts/context-hook.js:30
    ${u}`),u}var V=M.join(Mt(),".claude","plugins","marketplaces","thedotmack"),B=L(T.HEALTH_CHECK),h=null;function
    E(){if(h!==null)return h;try{let n=M.join(p.get("CLAUDE_MEM_DATA_DIR"),"settings.json"),t=p.loadFromFile(n);return
    h=parseInt(t.CLAUDE_MEM_WORKER_PORT,10),h}catch(n){return c.debug("SYSTEM","Failed to load port from settings, using
     default",{error:n}),h=parseInt(p.get("CLAUDE_MEM_WORKER_PORT"),10),h}}async function I(){try{let n=E();return(await
     fetch(`http://127.0.0.1:${n}/api/readiness`,{signal:AbortSignal.timeout(B)})).ok}catch(n){return
    c.debug("SYSTEM","Worker readiness check failed",{error:n instanceof
    Error?n.message:String(n),errorType:n?.constructor?.name}),!1}}function Pt(){try{let
    n=M.join(V,"package.json");return JSON.parse(Lt(n,"utf-8")).version}catch(n){return c.debug("SYSTEM","Failed to read
     plugin version",{error:n instanceof Error?n.message:String(n)}),null}}async function kt(){try{let n=E(),t=await
    fetch(`http://127.0.0.1:${n}/api/version`,{signal:AbortSignal.timeout(B)});return t.ok?(await
    t.json()).version:null}catch(n){return c.debug("SYSTEM","Failed to get worker version",{error:n instanceof
    Error?n.message:String(n)}),null}}async function K(){let n=Pt(),t=await kt();!n||!t||n!==t&&(c.info("SYSTEM","Worker
     version mismatch detected - restarting worker",{pluginVersion:n,workerVersion:t}),await new
    Promise(e=>setTimeout(e,L(T.PRE_RESTART_SETTLE_DELAY))),await A.restart(E()),await new
    Promise(e=>setTimeout(e,1e3)),await I()||c.error("SYSTEM","Worker failed to restart after version
    mismatch",{expectedVersion:n,runningVersion:t,port:E()}))}async function bt(){let
    n=p.get("CLAUDE_MEM_DATA_DIR"),t=M.join(n,".pm2-migrated");if(Rt(n,{recursive:!0}),!wt(t))try{Dt("pm2",["delete","cl
    aude-mem-worker"],{stdio:"ignore"}),j(t,new Date().toISOString(),"utf-8"),c.debug("SYSTEM","PM2 cleanup completed
    and marked")}catch{j(t,new Date().toISOString(),"utf-8")}let e=E(),r=await A.start(e);return
    r.success||c.error("SYSTEM","Failed to start
    worker",{platform:process.platform,port:e,error:r.error,marketplaceRoot:V}),r.success}async function G(){if(await
    I()){await K();return}if(!await bt()){let e=E();throw new Error(_({port:e,customPrefix:`Worker service failed to
    start on port ${e}.`}))}for(let e=0;e<5;e++)if(await new Promise(r=>setTimeout(r,500)),await I()){await
    K();return}let t=E();throw c.error("SYSTEM","Worker started but not responding to health checks"),new
    Error(_({port:t,customPrefix:`Worker service started but is not responding on port ${t}.`}))}function Y(n){throw
    n.cause?.code==="ECONNREFUSED"||n.code==="ConnectionRefused"||n.name==="TimeoutError"||n.message?.includes("fetch
    failed")||n.message?.includes("Unable to connect")?new Error(_()):n}function X(n,t,e){c.error("HOOK",`${e.operation}
     failed`,{status:n.status,...e},t);let r=e.toolName?`Failed ${e.operation} for ${e.toolName}:
    ${_()}`:`${e.operation} failed: ${_()}`;throw new Error(r)}import It from"path";function
    J(n){if(!n||n.trim()==="")return c.warn("PROJECT_NAME","Empty cwd provided, using
    fallback",{cwd:n}),"unknown-project";let
    t=It.basename(n);if(t==="")if(process.platform==="win32"&&n.match(/^[A-Z]:\\/i)){let
    o=`drive-${n[0].toUpperCase()}`;return c.info("PROJECT_NAME","Drive root detected",{cwd:n,projectName:o}),o}else
    return c.warn("PROJECT_NAME","Root directory detected, using fallback",{cwd:n}),"unknown-project";return t}async
    function q(n){await G();let t=n?.cwd??process.cwd(),e=J(t),r=E(),o=`http://127.0.0.1:${r}/api/context/inject?project
    =${encodeURIComponent(e)}`;try{let s=await fetch(o,{signal:AbortSignal.timeout(T.DEFAULT)});if(!s.ok){let a=await
    s.text();X(s,a,{hookName:"context",operation:"Context generation",project:e,port:r})}return(await
    s.text()).trim()}catch(s){Y(s)}}var yt=process.argv.includes("--colors");if(y.isTTY||yt)q(void
    0).then(n=>{console.log(n),process.exit(0)});else{let n="";y.on("data",t=>n+=t),y.on("end",async()=>{let
    t=n.trim()?JSON.parse(n):void 0,e=await q(t);console.log(JSON.stringify({hookSpecificOutput:{hookEventName:"SessionS
    tart",additionalContext:e}})),process.exit(0)})}

    Error: Worker service failed to start on port 37777. (port 37777)

    To restart the worker:
    1. Exit Claude Code completely
    2. Open Command Prompt or PowerShell
    3. Navigate to: %USERPROFILE%\.claude\plugins\marketplaces\thedotmack
    4. Run: npm run worker:restart
    5. Restart Claude Code
        at G (file:///C:/Users/BAIJU/.claude/plugins/marketplaces/thedotmack/plugin/scripts/context-hook.js:30:2154)
        at async q
    (file:///C:/Users/BAIJU/.claude/plugins/marketplaces/thedotmack/plugin/scripts/context-hook.js:30:3393)
        at async Socket.<anonymous>
    (file:///C:/Users/BAIJU/.claude/plugins/marketplaces/thedotmack/plugin/scripts/context-hook.js:30:3941)

    Node.js v22.15.0

Additional Context

Any other context about the problem.

[LoopOuroboros]

plugin version 7.3.9

[2025-12-18 08:59:38.507] [ERROR] [SYSTEM] Failed to start worker {platform=win32, port=37777, error=Process died during startup

[Suaig]

I have the same problem too.

[icew4y]

same problem here as well.

[iefnaf]

I also encountered this issue. It’s because Bun wasn’t installed when setting up claude-mem, which caused npm run worker:start to fail. You can first refer to https://bun.com/ to install Bun, then run npm run worker:restart.

[Ynohtna-Fr]

Still happening, even when I install bun previously

[AnemoneIndicum]

I also encountered this issue. It’s because Bun wasn’t installed when setting up claude-mem, which caused npm run worker:start to fail. You can first refer to https://bun.com/ to install Bun, then run npm run worker:restart.

I just installed bun, and the problem persists.

PS C:\Users\BAIJU\.claude\plugins\marketplaces\thedotmack> bun --version
1.3.5
PS C:\Users\BAIJU\.claude\plugins\marketplaces\thedotmack> npm run worker:restart

> claude-mem@7.4.0 worker:restart
> bun plugin/scripts/worker-cli.js restart

Failed to restart: Worker failed to start on Windows (readiness check timed out after 20000ms)

Troubleshooting:
1. Check Task Manager for zombie 'bun.exe' or 'node.exe' processes
2. Verify port 37777 is not in use: netstat -ano | findstr 37777
3. Check worker logs in ~/.claude-mem/logs/
4. See GitHub issues: #363, #367, #371, #373
5. Docs: https://docs.claude-mem.ai/troubleshooting/windows-issues

[lanyun1103]

Same error here. I have bun on my Win11

[ToxMox]

Is the port still in a zombie state from the previous version? Have you guys rebooted and tried again?

[ToxMox]

Just an FYI if the port is already in a zombie state attached to a closed process from a previous version of the plugin and you haven't rebooted yet, Nothing will kill the open port except a reboot. Nothing. I'm not able to replicate this issue anymore on the latest version since my wrapper method got added. If you're still getting the error after all of that then please tell me specifics of how your running claude code etc., I run the Windows native version of Claude Code (not the WSL version) from a Powershell 7 terminal either from Windows Terminal or in a powershell 7 terminal inside vs code in all my testing. (Also just tested using the Claude Code VS Code extension and that seems to work fine too for me.)

[techone-ekucukberber]

For now you can just give claude this workaround.

1. Stale PID file from dead worker process blocks restart
2. .ps1 file association may be broken (e.g., opening in Notepad instead of PowerShell)

Fix Steps:

1. Remove stale PID file:
   rm ~/.claude-mem/worker.pid

2. Start worker using direct bun.exe path (bypasses wrapper scripts):
   cd ~/.claude/plugins/marketplaces/thedotmack
   ~/.bun/bin/bun.exe plugin/scripts/worker-service.cjs

3. Permanent fix - add to PowerShell profile ($PROFILE):

Prioritize real bun.exe over npm wrapper scripts

$bunPath = "$env:USERPROFILE.bun\bin"
if (Test-Path $bunPath) {
$env:PATH = "$bunPath;$env:PATH"
}

4. Verify:
   curl http://127.0.0.1:37777/health

Should return: {"status":"ok"}

Alternatively, you could fix the .ps1 association: Right-click a .ps1 file → "Open with" → "Choose another app" → select
PowerShell → check "Always use this app".

This is because when I run claude code after I installed the claude mem plugin it asked me to choose an app to open the bun.ps1 file. I accidently chose notepad. Maybe you made a similar mistake too.

[LoopOuroboros]

For now you can just give claude this workaround.

1. Stale PID file from dead worker process blocks restart
2. .ps1 file association may be broken (e.g., opening in Notepad instead of PowerShell)

Fix Steps:

1. Remove stale PID file:
   rm ~/.claude-mem/worker.pid
2. Start worker using direct bun.exe path (bypasses wrapper scripts):
   cd ~/.claude/plugins/marketplaces/thedotmack
   ~/.bun/bin/bun.exe plugin/scripts/worker-service.cjs
3. Permanent fix - add to PowerShell profile ($PROFILE):

Prioritize real bun.exe over npm wrapper scripts

$bunPath = "$env:USERPROFILE.bun\bin" if (Test-Path $bunPath) { $env:PATH = "$bunPath;$env:PATH" }

4. Verify:
   curl http://127.0.0.1:37777/health

Should return: {"status":"ok"}

Alternatively, you could fix the .ps1 association: Right-click a .ps1 file → "Open with" → "Choose another app" → select PowerShell → check "Always use this app".

This is because when I run claude code after I installed the claude mem plugin it asked me to choose an app to open the bun.ps1 file. I accidently chose notepad. Maybe you made a similar mistake too.

This project has undergone so many iterations; it shouldn't fail to launch simply because bun is installed via npm .

[AnemoneIndicum]

For now you can just give claude this workaround.目前你可以直接给 claude 这个变通方法。

1. Stale PID file from dead worker process blocks restart过期的 PID 文件会阻止重启
2. .ps1 file association may be broken (e.g., opening in Notepad instead of PowerShell).ps1 文件关联可能已损坏（例如，用记事本打开而不是 PowerShell）

Fix Steps:  修复步骤：

1. Remove stale PID file:  移除过期的 PID 文件：
   rm ~/.claude-mem/worker.pid
2. Start worker using direct bun.exe path (bypasses wrapper scripts):使用直接的 bun.exe 路径启动 worker（绕过包装脚本）：
   cd ~/.claude/plugins/marketplaces/thedotmack
   ~/.bun/bin/bun.exe plugin/scripts/worker-service.cjs
3. Permanent fix - add to PowerShell profile ($PROFILE):永久修复方案 - 添加到 PowerShell 配置文件 ($PROFILE):

Prioritize real bun.exe over npm wrapper scripts优先使用真实的 bun.exe 而不是 npm 包装脚本

$bunPath = "$env:USERPROFILE.bun\bin" if (Test-Path $bunPath) { $env:PATH = "$bunPath;$env:PATH" }

4. Verify:  验证:
   curl http://127.0.0.1:37777/health

Should return: {"status":"ok"}应返回: {"status":"ok"}

Alternatively, you could fix the .ps1 association: Right-click a .ps1 file → "Open with" → "Choose another app" → select PowerShell → check "Always use this app".或者，您可以修复 .ps1 文件的关联：右键点击 .ps1 文件 → "打开方式" → "选择其他应用" → 选择 PowerShell → 勾选 "始终使用此应用"。
This is because when I run claude code after I installed the claude mem plugin it asked me to choose an app to open the bun.ps1 file. I accidently chose notepad. Maybe you made a similar mistake too.这是因为在我安装了 claude mem 插件后运行 claude 代码时，它提示我选择一个应用程序来打开 bun.ps1 文件。我错误地选择了 notepad。也许你也犯了类似的错误。

This project has undergone so many iterations; it shouldn't fail to launch simply because bun is installed via npm .这个项目已经经历了这么多迭代；它不应该仅仅因为通过 npm 安装了 bun 而无法启动。

Hello, my error is that after updating to version 7.3.9 or later, the service corresponding to port 37777 fails to start on the first attempt when launching the Claude Code CLI, even manually via bun or npm commands. Therefore, the corresponding ps1 file is missing.

PS C:\Users\BAIJU\.claude\plugins\marketplaces\thedotmack> ~/.bun/bin/bun.exe plugin/scripts/worker-service.cjs
[2025-12-19 10:30:20.193] [ERROR] [SYSTEM] ✗ Worker failed to start Failed to start server. Is port 37777 in use?
PS C:\Users\BAIJU\.claude\plugins\marketplaces\thedotmack> npm run worker:restart

> claude-mem@7.4.0 worker:restart
> bun plugin/scripts/worker-cli.js restart

Failed to restart: Worker failed to start on Windows (readiness check timed out after 20000ms)

Troubleshooting:
1. Check Task Manager for zombie 'bun.exe' or 'node.exe' processes
2. Verify port 37777 is not in use: netstat -ano | findstr 37777
3. Check worker logs in ~/.claude-mem/logs/
4. See GitHub issues: #363, #367, #371, #373
5. Docs: https://docs.claude-mem.ai/troubleshooting/windows-issues