MCP Toggl Server
# MCP Toggl
> Talk to your time tracking. Pull reports, start timers, inspect desktop activity, and turn raw Toggl data into useful recaps from Claude or any MCP-compatible client.
[](https://www.npmjs.com/package/@verygoodplugins/mcp-toggl)
[](https://modelcontextprotocol.io)
[](LICENSE)
[](https://developers.track.toggl.com/docs/)
## See What Your Week Actually Looked Like
> You ask: "Give me the recap of my week"
The MCP server returns hydrated time entries with project, workspace, client, tag, and running-timer context. Your client can turn that into a readable recap:
> 32.7 hours across 31 entries and 7 projects. Monday was a light planning day, Tuesday had the long implementation block, Wednesday stayed focused on one project, and Friday turned into a shipping run. The top project took 27% of the week, with two smaller projects close behind.

The server does not hard-code this prose or chart. It exposes structured Toggl data in a shape that makes synthesis easy.
## Catch the Drift
> You ask: "Did I actually work on what I said I worked on for that PR review entry?"
`toggl_get_timeline` can compare a tracked entry boundary with Toggl Track Desktop activity:
> The entry ran 1h 33m. About 67 minutes were in review tools, 5 minutes were scattered across chat apps, and the rest was idle or trimmed timeline space. The entry mostly checks out.

This is useful before invoicing, after long context-switching days, or whenever a vague entry like "admin" starts hiding too much Slack and browser time.
## See Patterns Over Time
> You ask: "Show me last month at a glance"
Daily and weekly report tools make it straightforward for the client to render heatmaps, spot streaks, and surface intensity changes:

Toggl is still the source of truth. The MCP layer makes the data easier for an agent to inspect, summarize, and visualize.
## Things You Can Ask
```text
What am I currently tracking?
How much time did I spend on the website project this month?
Start a timer for "PR review" on the Platform project
Show me yesterday's hours as a chart
What apps did I use most today?
Generate a daily report for last Friday
Compare this week to last week by project
Which day this month had the most billable work?
```
Chart prompts depend on your MCP client. The server returns the structured data; clients such as Claude decide how to render it.
## What Makes This Useful
**Hydrated responses**: time entries are enriched with `project_name`, `client_name`, `workspace_name`, `tag_names`, and normalized running-timer fields so the client does not need a second lookup for ordinary reporting.
**Smart caching**: workspaces, projects, clients, tasks, and tags are cached after first read. `toggl_cache_stats` shows hits, misses, loaded entities, and hit rate.
**Desktop activity timeline**: `toggl_get_timeline` summarizes app usage from Toggl Track Desktop and can return raw events when you need sequence analysis.
**Privacy controls**: timeline calls support summary-only output with `include_events: false` and title redaction with `redact_titles: true`.
**Period shortcuts**: `today`, `yesterday`, `week`, `lastWeek`, `month`, and `lastMonth` are supported on the tools where those periods make sense.
**Recoverable errors**: workspace resolution errors include `available_workspaces`, and Toggl quota/rate-limit errors include structured retry hints.
## Quick Start
### Prerequisites
- Node.js `^20.19.0` or `>=22.12.0`
- A Toggl Track account
- Your Toggl API token from [track.toggl.com/profile](https://track.toggl.com/profile)
### Claude Desktop
Add this to `~/Library/Application Support/Claude/claude_desktop_config.json`:
```json
{
"mcpServers": {
"mcp-toggl": {
"command": "npx",
"args": ["-y", "@verygoodplugins/mcp-toggl@latest"],
"env": {
"TOGGL_API_KEY": "your_api_key_here",
"TOGGL_DEFAULT_WORKSPACE_ID": "123456"
}
}
}
}
```
`TOGGL_DEFAULT_WORKSPACE_ID` is optional. If you have exactly one Toggl workspace, the server can resolve it automatically. If you have multiple workspaces and do not set a default, workspace-scoped tools return the available workspace IDs so the client can retry with `workspace_id`.
Restart Claude Desktop, then ask:
```text
What am I currently tracking?
```
### Global Install
```bash
npm install -g @verygoodplugins/mcp-toggl
mcp-toggl --help
```
## Tools
### Reports and Insights
| Tool | What it does |
| --- | --- |
| `toggl_daily_report` | Hours by project and workspace for a date. Use `format: "text"` for display text or `"json"` for structured output. |
| `toggl_weekly_report` | 7-day breakdown with daily totals and project rollups. Use `week_offset: -1` for last week. |
| `toggl_get_time_entries` | Raw hydrated entries by period, date range, workspace, or project. |
| `toggl_get_timeline` | Toggl Track Desktop app usage summary with optional raw events. |
### Timer Control
| Tool | What it does |
| --- | --- |
| `toggl_get_current_entry` | Returns the running timer, elapsed seconds, and hydrated project/workspace context. |
| `toggl_start_timer` | Starts a timer with description, optional project/task, and tags. |
| `toggl_stop_timer` | Stops the currently running timer. |
### Lookups
| Tool | What it does |
| --- | --- |
| `toggl_check_auth` | Verifies token access and lists available workspaces without exposing the token. |
| `toggl_list_workspaces` | Lists all accessible workspaces. |
| `toggl_list_projects` | Lists projects for a workspace using cache-backed reads after first fetch. |
| `toggl_list_clients` | Lists clients for a workspace using cache-backed reads after first fetch. |
### Cache Management
| Tool | What it does |
| --- | --- |
| `toggl_warm_cache` | Pre-fetches workspace, project, client, and tag data before a heavy reporting session. |
| `toggl_cache_stats` | Returns hits, misses, hit rate, loaded entity counts, and warm-cache state. |
| `toggl_clear_cache` | Clears cached data. Useful after creating or renaming Toggl entities. |
### Summaries
| Tool | What it does |
| --- | --- |
| `toggl_project_summary` | Total hours per project for a period or date range. |
| `toggl_workspace_summary` | Total hours per workspace for a period or date range. |
## Timeline Privacy
Toggl Track Desktop activity can include window titles. Those titles may contain document names, email subjects, chat text, URLs, OAuth pages, or database names.
Summary-only mode returns app totals without raw events:
```json
{
"period": "today",
"include_events": false
}
```
Events with redacted titles preserve sequence and duration but remove titles:
```json
{
"period": "today",
"redact_titles": true,
"limit": 50
}
```
Full event mode is the default:
```json
{
"period": "today"
}
```
When in doubt, use `include_events: false`.
## Configuration Reference
| Env var | Required | Default | Notes |
| --- | --- | --- | --- |
| `TOGGL_API_KEY` | Yes | - | Preferred env var for your Toggl API token. |
| `TOGGL_API_TOKEN` | No | - | Supported alias for backwards compatibility. `TOGGL_API_KEY` is preferred. |
| `TOGGL_TOKEN` | No | - | Supported alias for backwards compatibility. `TOGGL_API_KEY` is preferred. |
| `TOGGL_DEFAULT_WORKSPACE_ID` | No | - | Used when a tool requires a workspace and none is passed. |
| `TOGGL_CACHE_TTL` | No | `3600000` | Cache TTL in milliseconds. Default is 1 hour. |
| `TOGGL_CACHE_SIZE` | No | `1000` | Maximum cached entity budget. |
| `TOGGL_BATCH_SIZE` | No | `100` | Batch size used by API pagination helpers. |
## Caveats
**Toggl rate limits and quotas**: Toggl may return rate-limit or quota errors during chatty sessions. The server returns structured retry information when Toggl provides it. Warm the cache before large reporting sessions to avoid repeated project/client/tag fetches.
**Running timer duration**: Toggl uses negative duration values for running entries. Read `running` and `elapsed_seconds` from the hydrated response instead.
**Timeline availability**: `toggl_get_timeline` requires Toggl Track Desktop timeline sync. If it is not enabled or has not uploaded data yet, the tool returns `enabled: false` with setup guidance.
**Timeline totals**: `limit` only limits returned raw events. `summary`, `total_seconds`, and `total_hours` are calculated from all matching events.
## Local Development
```bash
git clone https://github.com/verygoodplugins/mcp-toggl.git
cd mcp-toggl
npm install
npm run build
npm test
```
Useful commands:
```bash
npm run dev
npm run lint
npm run format
```
## License
MIT.
Built by [Very Good Plugins](https://verygoodplugins.com).
TDQS
Scored across 15 tools
Every tool has a clearly distinct purpose with no ambiguity. Tools like toggl_start_timer and toggl_stop_timer handle timer control, while toggl_get_time_entries and toggl_get_current_entry focus on retrieving time data, and reporting tools like toggl_daily_report and toggl_weekly_report serve specific output formats. There is no overlap in functionality.
All tools follow a consistent verb_noun pattern with the 'toggl_' prefix, using snake_case throughout. Examples include toggl_list_workspaces, toggl_start_timer, and toggl_clear_cache, making the naming predictable and easy to understand.
With 15 tools, the count is well-scoped for a Toggl time-tracking server. It covers authentication, data listing, timer operations, reporting, and caching without being overwhelming, with each tool earning its place in the workflow.
The tool set provides strong coverage for core time-tracking operations, including CRUD-like actions (list, start/stop, get entries) and reporting. A minor gap is the lack of tools for updating or deleting time entries, which agents might need to work around, but overall, the surface is nearly complete for the domain.