safe-mind-tools
# Safe Mind Developer Tools
Official CLI and MCP server for [Safe Mind](https://www.safe-mind.de) - the platform
for the legally required psychological risk assessment (Gefährdungsbeurteilung
psychischer Belastung, GBU Psyche) in Germany.
| Package | npm | Purpose |
| --- | --- | --- |
| [`cli/`](./cli) | [safe-mind](https://www.npmjs.com/package/safe-mind) | Terminal client for the Safe Mind Public API v1 |
| [`mcp/`](./mcp) | [safe-mind-mcp](https://www.npmjs.com/package/safe-mind-mcp) | MCP server for AI assistants (Claude, ChatGPT, Gemini, Cursor, ...) |
## Quick start
```bash
npm install -g safe-mind
safe-mind login --email you@example.com --password '...'
safe-mind surveys create --start 2026-10-01 --weeks 4 --employees 220 \
--department "Produktion" --department "Verwaltung"
```
MCP (any MCP client):
```json
{
"mcpServers": {
"safe-mind": {
"command": "npx",
"args": ["-y", "safe-mind-mcp"],
"env": { "SAFE_MIND_API_KEY": "sm_live_..." }
}
}
}
```
## Documentation
- API reference and examples: https://www.safe-mind.de/developers
- Anonymity: analytics endpoints only return aggregated results; groups below the
minimum group size are filtered server-side.
## License
MIT
TDQS
Scored across 12 tools
Each tool targets a distinct resource+action pair. Departments and locations are clearly separated (list/create variants), surveys vs. analytics are distinct, and the account-lifecycle tools (register, verify_email, get_api_key, me) each have a unique role. No two tools could reasonably be confused.
All tools use a consistent snake_case safe_mind_ prefix followed by a predictable verb_noun pattern (list_, create_, get_, verify_). The convention is applied uniformly across the entire set.
At 12 tools the set is well-scoped for the domain, covering account setup, reference data, surveys, and analytics without redundancy. Every tool appears to earn its place.
Core workflows (register, get key, create departments/locations, create survey, view results) are covered, but there are notable gaps: departments and locations support create+list only with no update or delete, and survey drafts cannot be edited or removed via the API. Agents may hit dead ends when cleanup or correction is needed.