MCP HTTP endpoint for Nixie Flow diagram management system. Exposes tools for creating, reading, updating, and deleting Mermaid diagrams with grounding-based verification of code contracts embedded in diagram notes.
Nixie Flow exposes 17 tools with HTTP transport, but definition quality is significantly below production baseline. Most tools lack proper parameter descriptions (only 'name' and 'arguments' are documented; nested object structures are opaque). Tool descriptions are present but generic (e.g., 'Calls a specified MCP tool' for tools/call). No output schemas are documented. Critical tools like prepare_save and commit_save lack clear guidance on grounding verdicts structure. Error handling is absent from descriptions. The server conflates MCP protocol tools (initialize, ping, tools/list) with domain tools (diagram operations), violating single-responsibility principle.
Commits a prepared save with grounding verdicts
Creates a new diagram with optional slug and source
Soft deletes a diagram (owner only)
Retrieves diagram source, grounding, and revision_id (no layout)
Retrieves diagram layout (positions sidecar only)
Advertises tools and prompts; returns server instructions for grounding procedure
Lists owned and shared diagrams, excluding deleted ones
Opaque 'arguments' parameter: tools/call, prompts/get, and others accept an 'arguments' object with no schema or nested field documentation. LLMs cannot infer what keys/values are valid.
No output schemas documented. Callers cannot know what fields to expect from list_diagrams, get_diagram, prepare_save, etc. Breaks downstream tool chaining.
Protocol tools (initialize, ping, tools/list, prompts/list) mixed with domain tools (diagram operations). Violates single-responsibility; confuses LLM tool selection.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
Ping endpoint for connection verification
Prepares a save operation and returns token + requires_grounding flag for grounding gate
Retrieves the grounding prompt with slug substitution
Lists available MCP prompts
Yields the edit turn (scepter) to allow other editors to proceed
Saves diagram source (DEPRECATED - use prepare_save/commit_save instead)
Re-verifies diagram in-place without snapshot (source unchanged)
Updates diagram layout with expected version for conflict detection
Calls a specified MCP tool with provided arguments
Lists all available MCP tools
Grounding verdicts structure undefined. prepare_save returns 'requires_grounding' flag; commit_save accepts 'grounding' object. No schema for verdict keys/values.
No error handling guidance. Descriptions do not explain what errors can occur, whether they are retryable, or what the LLM should do next.