prompt-time

prompt-time

Schedule Windows desktop reminders from Claude Desktop using PowerShell and Windows Task Scheduler. Supports one-time and recurring reminders that persist across reboots.

Category
Visit Server

README

prompt-time

CI

Schedule Windows desktop reminders directly from Claude Desktop. Zero dependencies — just PowerShell + Windows Task Scheduler.

What it does

Adds three tools to Claude Desktop:

Tool What it does
schedule_reminder Set a one-time or recurring desktop reminder
list_reminders Show every reminder currently scheduled
cancel_reminder Cancel a reminder by ID

Reminders persist across reboots and Claude restarts. There is no cloud surface — everything runs locally.

Install in 60 seconds

  1. Download prompt-time-v2.2.6.zip from the latest release.
  2. Right-click the zip → Extract All. Windows must extract — running install.bat from inside the zip preview window will not work.
  3. Open the extracted folder, double-click install.bat.
  4. If Claude Desktop is running, the installer will prompt to restart it. Accept — the MCP tools only show up after Claude Desktop is restarted.
  5. Done. The tools are live.

No Node, no npm, no admin rights, no internet required.

The extracted folder is disposable after install. v2.2.6+ copies both the MCP server and the watcher into the prompt-time data directory at install time, and points Claude Desktop's config + the scheduled task at the data-dir copies. You can move or delete the extracted folder afterwards without breaking anything. (Re-run install.bat from a fresh extract if you ever want to upgrade.)

Usage examples

Just talk to Claude:

"Remind me to send the weekly update at 4pm today" "Set a reminder every weekday at 9am to check my inbox" "Remind me about the board call next Monday at 2pm" "What reminders do I have set?" "Cancel reminder PROMPTTIME-A1B2C3D4"

Recurrence: once, daily, weekly, weekdays. Granularity is ~10 seconds (the watcher's poll interval).

How it works

Two pieces:

  1. MCP server (prompt_time.ps1) — runs inside Claude Desktop's process tree. Validates and enqueues reminders to a JSONL file on disk.
  2. Watcher daemon (prompt-time-watcher.ps1) — registered once at install as a Task Scheduler At LogOn task. Polls the queue, renders the toast popup in-process, and advances recurring entries.

The MCP server never talks to Task Scheduler at runtime, never spawns a process, and never displays UI. It only writes one JSONL line per reminder. Everything else is the watcher's job.

Why two processes — the MSIX boundary

Claude Desktop ships as an MSIX-packaged app. MSIX containers transparently virtualize file writes: a server running inside the package that calls Add-Content $env:APPDATA\prompt-time\queue.jsonl does not write to %APPDATA%\Roaming\prompt-time\queue.jsonl — it writes to a per-package shadow tree under %LOCALAPPDATA%\Packages\<package-family>\LocalCache\….

Anything Task Scheduler spawns runs outside the package, so a per-reminder schtasks task would look for files at the real path and find nothing.

prompt-time resolves this once: at startup, every component (prompt_time.ps1, prompt-time-watcher.ps1, install.ps1, uninstall.ps1) discovers the Claude AppX package and computes the same canonical data directory inside its LocalCache. Server writes and watcher reads land at the same physical file. The watcher renders the popup itself rather than asking Task Scheduler to do it, which keeps every state-bearing operation on the same side of the boundary.

The installer also pre-creates the queue file at the canonical path before the MCP server's first write — without that, MSIX file-virtualization redirects new file creation into the package shadow even though existing files at the real path are written through.

The same trap bites the Claude Desktop config itself. If MSIX-packaged Claude has ever written to its own claude_desktop_config.json (e.g. saving preferences), MSIX materializes a per-package shadow copy at %LOCALAPPDATA%\Packages\<family>\LocalCache\Roaming\Claude\claude_desktop_config.json and then reads from the shadow on every subsequent boot. An installer that writes only to %APPDATA% becomes invisible to Claude on those machines. v2.2.3+ reads shadow-first as the source of truth and writes the merged config to both paths atomically; uninstall removes from both. v2.2.4 adds post-write self-verify so install fails loudly with an actionable error if the entry doesn't survive (AV rollback, managed-policy override, etc.) instead of reporting a false success. Run diagnose.bat for a full breakdown.

Troubleshooting

Notifications muted by Focus Assist. Windows Focus Assist (Settings → System → Focus) silences toasts. Either turn it off or add powershell.exe to the Priority list.

Tools don't appear in Claude Desktop.

  1. Fully quit Claude Desktop (right-click tray icon → Quit) and reopen it. The "X" close button only minimizes; tools won't reload.
  2. Run diagnose.bat from the extracted folder. It reads your config, spawns prompt_time.ps1 the way Claude Desktop will, reads Claude Desktop's MCP logs, and prints an actionable hint for each failure. Most "install ran but nothing appeared" cases are diagnosed here without needing IT.
  3. If diagnose.bat says all checks pass but tools still don't show: run diagnose.bat -Json and attach the output when filing a bug.

install.bat closes immediately. Run it from a cmd prompt instead so you can read the error: install.bat from inside the unzipped folder. The installer also writes a debug log at %APPDATA%\prompt-time\prompt-time.debug.log (or the MSIX LocalCache equivalent) — check there.

Reminders not firing. Check the watcher task is running:

Get-ScheduledTask -TaskName PROMPTTIME-Watcher | Get-ScheduledTaskInfo

LastTaskResult should be 267009 (running) or 0 (success).

Uninstall

Double-click uninstall.bat. It removes the prompt-time entry from Claude Desktop's config, deletes the PROMPTTIME-Watcher scheduled task, and clears the data directory. Other MCP servers in your config are untouched.

Tests

.\tests\run.ps1

Installs Pester 5 if missing, runs the Pester suite, and emits NUnit XML at tests/pester-results.xml. CI runs the same harness plus Invoke-ScriptAnalyzer on every push.

Security

See SECURITY.md for the trust model, supply-chain story, and how to report a vulnerability. Short version: the trust boundary is your local Windows user; reminder titles and messages are stored on disk in cleartext (don't put secrets in them); prompt-time registers a logon-time scheduled task as a standard persistence primitive.

License

MIT — see LICENSE.


Built by Ryan Stewart

Recommended Servers

playwright-mcp

playwright-mcp

A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.

Official
Featured
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.

Official
Featured
Local
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.

Official
Featured
Local
TypeScript
VeyraX MCP

VeyraX MCP

Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.

Official
Featured
Local
graphlit-mcp-server

graphlit-mcp-server

The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.

Official
Featured
TypeScript
Kagi MCP Server

Kagi MCP Server

An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.

Official
Featured
Python
Neon Database

Neon Database

MCP server for interacting with Neon Management API and databases

Official
Featured
E2B

E2B

Using MCP to run code via e2b.

Official
Featured
Exa Search

Exa Search

A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.

Official
Featured
Qdrant Server

Qdrant Server

This repository is an example of how to create a MCP server for Qdrant, a vector search engine.

Official
Featured