MCP server for DevDash development - explorer interaction, screenshots, telemetry, and logs
DevDash MCP has 16 tools with reasonable naming conventions and descriptions, but exhibits significant gaps in schema completeness, parameter documentation, and error handling guidance. Tool names follow verb_noun patterns (qml_explorer_*, screenshot_*, devdash_*) which is good. Descriptions are present and generally adequate (100-300 chars), meeting the 10-1024 character baseline. However, critical issues emerge: (1) Most tools declare input as {} (empty), making schema validation impossible for LLMs. (2) Parameter descriptions are minimal or absent for critical fields like 'window' in screenshot tools. (3) No documented output schemas, LLMs cannot predict response structure for composition. (4) Error handling is minimal; tools don't guide recovery or categorize failures. (5) STDIO transport caps overall readiness. The server is functional for a development tool but lacks production-grade definition quality.
Retrieve recent application logs from DevDash (devdash repo). Fetches logs from the DevDash application's internal log buffer. Useful for debugging adapter issues, CAN parsing errors, or configuration problems. Requires DevDash to be running with DevTools enabled.
Get current telemetry state from DevDash (devdash repo). Returns real-time vehicle data including RPM, speed, temperatures, pressures, and other sensor values from the DataBroker. Requires DevDash to be running with DevTools enabled.
Get active warnings and critical alerts from DevDash (devdash repo). Returns any active warning conditions such as high temperature, low oil pressure, or other alert states. Requires DevDash to be running with DevTools enabled.
List DevDash windows available for screenshot (devdash repo). Uses the DevTools HTTP API to list windows, which may differ from X11 window detection.
Capture screenshot via DevDash DevTools API (devdash repo). Captures screenshots through the DevDash application itself, which may provide higher quality than X11-based capture.
Missing input schemas (12 of 16 tools declare input as {}). LLMs cannot validate parameters or understand what fields are accepted. Tools like qml_explorer_status, qml_explorer_get_state, qml_explorer_list_properties, screenshot_list_windows, devdash_telemetry_get_state, devdash_telemetry_get_warnings, devdash_telemetry_list_windows, qml_explorer_build, and qml_explorer_launch all have zero schema definition.
No documented output schemas. Tools do not declare what fields or structure they return. LLMs cannot plan downstream tool compositions (e.g., what ID field to pass to the next tool). All tools lack return type documentation.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 40 | - | v1 |
Build the QML Gauges Explorer (qml-gauges repo). Runs cmake configure and build for the explorer. This is required before launching the explorer if the code has changed. Requires DEVDASH_QML_GAUGES_PATH to be set in .env or environment.
Get the current value of a specific property in the QML Gauges Explorer (qml-gauges repo).
Get the current state of the QML Gauges Explorer (qml-gauges repo). Returns the current page, available properties, and their values. Requires the explorer to be running.
Launch the QML Gauges Explorer (qml-gauges repo). Starts the explorer with the correct library paths. The explorer provides a WebSocket server on port 9876 for property inspection and modification. Requires DEVDASH_QML_GAUGES_PATH to be set in .env or environment.
List all available properties for the current component page in QML Gauges Explorer (qml-gauges repo). Returns full documentation for each property including name, type, range, default value, and description.
Navigate the QML Gauges Explorer to a specific component page (qml-gauges repo).
Set a property value on the current component in the QML Gauges Explorer (qml-gauges repo).
Check if the QML Gauges Explorer is running (qml-gauges repo). Returns the running status, process IDs if running, and whether the WebSocket server is responding.
Capture a PNG screenshot of a window.
Capture a compact screenshot focused on the gauge preview area. IMPORTANT: This is the primary screenshot tool for verifying gauge visuals. It crops to the LEFT 60% (preview pane) then focuses on the center where the gauge is rendered. The explorer layout is: - Left 60%: Preview area with gauge centered - Right 40%: Property panel (excluded from capture) Uses ~5k tokens instead of ~25k for full screenshots.
List available windows for screenshot capture. Returns windows matching DevDash-related keywords (gauge, explorer, devdash, qml, cluster, headunit).
Parameter descriptions are minimal or absent for key fields. 'window' parameter in screenshot tools (screenshot_capture, screenshot_gauge_preview, devdash_telemetry_screenshot) has brief descriptions but lacks examples of valid values or substring matching behavior. 'value' in qml_explorer_set_property states 'type depends on property' without guidance on discovering valid values.
No error handling guidance. Tools do not document failure modes or what the LLM should do on error. For example, qml_explorer_navigate says 'Requires the explorer to be running' but does not provide a recovery suggestion if navigation fails. No error classification (retryable vs fatal).
Tool names contain domain prefix + action, which is verbose and repetitive. qml_explorer_* and devdash_telemetry_* prefixes add verbosity without improving clarity. In a tool list, these names could be shortened (e.g., explorer_navigate, telemetry_get_state) with organization via tool categories/tags.
Ambiguous resource handling. screenshot_capture and devdash_telemetry_screenshot both capture screenshots but via different backends (X11 vs DevTools API). Tool descriptions do not explicitly state when to use one vs the other, forcing LLM reasoning.
Missing dependencies not declared. qml_explorer_build and qml_explorer_launch require DEVDASH_QML_GAUGES_PATH environment variable. If missing, failures are opaque. No validation or user guidance in tool descriptions.