SuperAPI AI Directory

superglookoquery

Clinical audit of Glooko diabetes data from any pump or CGM: time in range, GMI, trends and bolus/basal analysis, adapting to the data your devices report.

Open website

About this tool

SuperGlookoQuery is an MCP extension running in Claude Desktop for clinical auditing of Glooko diabetes device data. It supports any pump and CGM combination, and through a device-agnostic capability discovery mechanism, it dynamically builds the tool list based on the data fields actually present in the user's account, rather than assuming the field set of a fixed device.

Purpose Users can directly ask Claude about their own diabetes data in natural language, for example "How was my time in range last month", "Why am I always high at night", "Show me my worst day and tell me what happened". The extension invokes analysis tools in real time and provides answers within the conversation.

Main capabilities

  • On-demand summaries: Time in Range, GMI, variability, best/worst dates and hours, basal/bolus insulin distribution, for any specified time window.
  • Locally saves full-resolution CGM data: every 5-minute reading is archived, but Claude prioritizes aggregated results and only pulls raw readings when a question truly requires them.
  • Real charts: opens clinical report-style glucose charts in the browser, including color-coded ranges, hoverable bolus insulin markers, and day-by-day breakdowns.
  • Bolus insulin context: each bolus is matched with the ISF, carb ratio, and target in effect at that time, including split or extended boluses reported by the pump, making it easier to judge whether the dose was reasonable.
  • Basal insulin status: presented as states such as suspended, running at maximum, or blind spot after signal loss, rather than raw infusion units.
  • Device-specific extensions: for example the CamAPS FX operating mode breakdown, which appears only after the account data actually produces that field.

How to use This extension can only run in Claude Desktop (macOS or Windows), and is not available for web or mobile Claude. Use requires Claude Desktop and (when analyzing your own data) a Glooko account with a synced device. No Docker, standalone Node installation, or terminal is required. Installation: obtain the .mcpb file from the repository Releases page (if published), or build it yourself. Installation methods include double-clicking the .mcpb file, dragging it into the Claude Desktop window, or via Settings → Extensions → Advanced settings → Install extension. After installation, review permissions and confirm. Configuration: Claude Desktop automatically generates the settings form; no .env file. Most fields have sensible defaults. The Glooko email and password are the only optional pair; leaving both blank enters offline sample data mode. When switching the display unit to mgdl, the low/high thresholds must be updated at the same time (default 3.9/10.0 mmol/L, no automatic conversion). Connecting an account: enter the Glooko email and password, set the account blood glucose unit to match the real Glooko setting, and choose the display unit and target range. The first question afterward triggers a one-time historical download (a few seconds to about a minute), after which the local archive is read. Usage: start a conversation in Claude Desktop, confirm the extension is enabled, then ask a question, for example "tell me about my diabetes data". You can continue asking follow-up questions, drill into a specific day, or request charts. Clinical audit persona: request "use the clinical audit persona" or "just give me the full audit", and Claude will call the activate_clinical_auditor_persona tool and adopt that persona.

Applicable scenarios Suitable for users who sync pump or CGM data to Glooko for personal data understanding and appointment preparation. It has currently been built and tested end-to-end on one real configuration: Ypsomed YpsoPump running CamAPS FX as the AID app, paired with Libre 3+ CGM. Other pump/CGM combinations require real account contributions of discovery reports before they can receive actual support. Without a Glooko account, you can also use the built-in fully synthetic sample data to explore offline.

Limitations

  • Not medical advice, not a medical device, and should not be used on its own to change treatment plans.
  • Claude Desktop only.
  • Limited device support, which needs to be gradually expanded with real account data.
  • Claude Desktop currently has no menu for selecting MCP prompts, so the clinical audit persona must be requested in natural language.
  • It involves real medical credentials and health data; credentials are stored in Claude Desktop's secure settings; when discussing data, it is recommended to turn off chat history or model training.
  • The data contribution process includes mandatory human review, a confirmation phrase, and a second privacy scan, and does not include real values.