Vetroscope MCP
A read-only Model Context Protocol server for Vetroscope that provides LLMs with detailed time-tracking data, including reports, app/project breakdowns, goals progress, and activity status, all from a local SQLite database.
README
Vetroscope MCP
A read-only Model Context Protocol server for Vetroscope — gives LLMs (Claude Desktop, Claude Code, ChatGPT, Cursor, …) context on how you've been spending your time, what projects you've been in, and how you're tracking against your goals.
Reads your local Vetroscope SQLite directly, read-only. No cloud round-trip, no auth, works offline.
Tools
Reports & breakdowns
| Tool | What it does |
|---|---|
get_report |
Aggregate report for a period — total active seconds, top apps, top projects with sub-projects nested (YouTube videos, SoundCloud songs, Netflix episodes). |
get_app_breakdown |
Per-project breakdown for a single app over a period, with sub-projects. |
get_app_stats |
Deep stats for one app: lifetime totals, days active, daily series, hour-of-day distribution, weekday distribution. |
get_tag_breakdown |
Time-spent report for a single tag — top apps, top projects, daily series, active/passive split. |
get_calendar |
Dense per-day series (heatmap data) for any period — defaults to a full year. |
get_device_breakdown |
Per-device totals when you run Vetroscope across multiple machines. |
get_music_split |
Music-vs-work analysis: work-with-music / music-only / heads-down-work / other, plus per-source (Spotify, SoundCloud, …) overlap totals. Classifier is overridable per call. |
get_category_breakdown |
Time rolled up into broad categories: editor, browser, adobe, communication, gaming, productivity, creative, etc. Mirrors Vetroscope's internal app grouping. |
get_listening_history |
Top tracks and top artists across native music apps and browser music sites, plus per-day listening minutes. Artists parsed from the "Artist — Title" sub_project convention. |
get_media_links |
Canonical deep-links Vetroscope captured for media you actually played — Spotify spotify:track:… URIs and YouTube youtube.com/watch?v=… URLs — joined with the matching time data. Requires Vetroscope ≥ 0.2.30 with capture_media_links enabled. |
get_focus_heatmap |
7×24 grid of active foreground seconds — when do you usually do specific kinds of work. Optional app / project / tag filter to narrow to a single activity. |
Reference / lookup
| Tool | What it does |
|---|---|
list_tags |
All your tags with id, name, color, sticky flag. |
list_projects |
Every (app, project) pair ever tracked with all-time totals + first/last seen + optional substring search. |
list_markers |
Your timeline markers (timestamp, label, color, icon, optional region end). |
Activity / status
| Tool | What it does |
|---|---|
get_sessions |
Continuous activity blocks (start/end/duration) — the natural grain for "what did I work on this morning". |
get_current_status |
Most recent entry — what app/project right now, how recently, tracking vs idle. |
query_entries |
Filtered list of raw 30s entries (app / project / tag / search / period / mode). |
Goals
| Tool | What it does |
|---|---|
get_goals_progress |
Current progress on configured app / overall / tag goals. |
get_goal_achievements |
Historical record of which goals you hit on which days — drives streak questions. |
All time-aware tools accept these shared filters:
| Argument | Purpose |
|---|---|
period |
today · yesterday · week · month · year · single date YYYY-MM-DD · inclusive range YYYY-MM-DD..YYYY-MM-DD |
hour_start / hour_end |
Inclusive / exclusive hour-of-day filter in local time (e.g. 9 and 17 = working hours) |
weekdays |
Array of weekday integers (0=Sun, 1=Mon, …, 6=Sat). Pass [1,2,3,4,5] for weekdays only. |
device |
current (the local machine), a device UUID from get_device_breakdown, or a platform name (darwin, win32, browser-extension). Omit for all devices. |
Every total is split into active foreground time and passive away-listening time (e.g. background music while idle), matching the dashboard's distinction.
Requirements
- Vetroscope installed and run at least once.
- Node.js 18+.
Install
npx vetroscope-mcp
That's it — no global install needed. The first run will fetch the package from npm.
Configure your client
Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"vetroscope": {
"command": "npx",
"args": ["-y", "vetroscope-mcp"]
}
}
}
Restart Claude Desktop.
Claude Code
claude mcp add vetroscope -- npx -y vetroscope-mcp
ChatGPT (Developer Mode)
In the Connectors section, add a new MCP server with command npx and args -y vetroscope-mcp.
Cursor
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"vetroscope": {
"command": "npx",
"args": ["-y", "vetroscope-mcp"]
}
}
}
How it finds your database
Vetroscope stores its SQLite at:
- macOS:
~/Library/Application Support/Vetroscope/ - Windows:
%APPDATA%\Vetroscope\ - Linux:
~/.config/Vetroscope/
If you're signed into a Vetroscope account, the active DB is vetroscope-<userId>.db (recorded in auth-state.json). Otherwise it's the anonymous vetroscope.db.
You can override either piece with environment variables:
| Env var | Purpose |
|---|---|
VETROSCOPE_DIR |
Override the app-data directory |
VETROSCOPE_DB_PATH |
Point at an explicit .db file |
Example prompts
- "How much time did I spend in After Effects this week?"
- "What did I work on yesterday?"
- "Am I on track to hit my coding goal today?"
- "What were the top three projects I touched this month?"
- "Show me every session that touched the 'Vetroscope' project this week."
- "How much time did I spend on the Vetroscope Dev tag this month?"
- "When during the day do I usually use Cursor?"
- "How many hours of focused work did I do during weekday working hours (9–5) last week?"
- "Which YouTube videos did I watch yesterday and how long?"
- "What am I doing right now?"
- "Show me my longest focused work sessions today."
- "How many days in a row have I hit my coding goal?"
- "What was happening during my 'Eye appointment' marker on Wednesday?"
- "Have I ever worked on a project called 'Atlas'?"
- "What was my busiest day this year?"
- "How much of my coding time was on my Mac vs Windows this month?"
- "How much of my work this week had music playing?"
- "How long did I spend just listening to music with nothing else open today?"
- "How much creative work vs coding did I do this month?"
- "Who were my top five artists this week?"
- "When during the week do I usually code in Cursor?"
Local development
git clone https://github.com/rankin-works/Vetroscope-MCP.git
cd Vetroscope-MCP
npm install
npm run build
node dist/index.js # starts a stdio MCP server (mainly useful via a client)
Type-check only:
npm run typecheck
How it works
The server is a thin query adapter over the same SQLite database that Vetroscope writes. It mirrors the bucket-distinct duration math used in electron/database.ts so totals match the desktop dashboard exactly.
Because it reads the schema directly, a Vetroscope schema migration could break the MCP. The set of tools is intentionally narrow — purpose-built rather than a generic SQL surface — so changes localize to src/queries.ts.
Stability guarantees
Starting with 1.0.0, vetroscope-mcp follows Semantic Versioning. The public API surface — tool names, parameter names, parameter semantics, and response field names — is stable and changes follow these rules:
| Change | SemVer bump |
|---|---|
| Adding a new tool | minor (1.x.0) |
| Adding a new optional argument to an existing tool | minor |
| Adding a new field to an existing response | minor |
| Renaming a tool, argument, or response field | major (2.0.0) |
| Changing a parameter's accepted values or default | major |
| Changing the meaning of an existing field (e.g. seconds → minutes) | major |
| Removing a tool, argument, or response field | major |
| Tightening validation in a way that breaks previously valid input | major |
Bug fixes and internal refactors that don't change the API surface are patch bumps (1.0.x).
Things explicitly not under the SemVer contract:
- Internal helpers and types not exported from the npm package
- The set of canonical app names in
src/categories.ts(thecategorizeAppfunction may classify the same app differently between minor versions as Vetroscope adds apps) - The exact wording of tool descriptions
- Error message text
If you build something on top of vetroscope-mcp and want to lock to a major version, pin to ^1.0.0 in your dependency.
See CHANGELOG.md for the full release history.
License
MIT © Jacob Rankin
Recommended Servers
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.
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.
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.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
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.
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.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
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.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.