HTTP MCP Transport for Nchan - A library for implementing MCP (Machine Conversation Protocol) over HTTP using Nchan with support for both custom tools and OpenAPI integration
This MCP server exhibits critical definition quality gaps that prevent reliable agent usage. Tool naming is inconsistent and non-descriptive (e.g., two tools named 'add' with different input types). Descriptions are minimal or missing context. Input schemas are present but parameter descriptions are sparse. The server appears to be a transport/infrastructure project (HTTP MCP Transport for Nchan) rather than a feature-complete tool server. The duplicate 'add' tool creates ambiguity that LLMs cannot resolve. Output schema documentation is entirely absent. Error handling guidance is not evident in the code samples provided.
Addition tool that adds two numbers
Adds two integers
Returns a hello world message
Long running task that sleeps for specified duration
Duplicate tool names: Two tools named 'add' with different input schemas (number vs integer). LLMs cannot distinguish between them and will make arbitrary selection errors.
Tool names lack action verbs. 'add' and 'hello_world' do not follow verb_noun convention. Should be 'sum_numbers' or 'calculate_sum'; 'greet_user' for hello_world. LLMs infer intent from names alone.
Minimal descriptions. 'Addition tool that adds two numbers' (33 chars) and 'Returns a hello world message' (30 chars) are below the 50-char minimum for actionable descriptions. No guidance on when to call vs alternatives, no expected output format details.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 42 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 31 | - | v1 |
Missing parameter descriptions. In typescript/server.js, parameters 'a' and 'b' have type info but lack descriptions explaining what they represent or constraints. Only inline type hints visible in schema, not in documentation.
No output schema documentation. Tools return content with 'type: text' wrapper but no documented return structure is visible. LLMs cannot predict response fields for downstream planning.
No error handling or recovery guidance in visible code. If 'long_task' timeout occurs or invalid sleep duration passed, no guidance provided to LLM on retry strategy or parameter constraints.
'long_task' parameter 't' lacks description. The parameter name 't' is cryptic; description says 'Sleep duration in seconds' but no range constraints documented. LLMs may pass invalid values (negative, extremely large).