MCP server for searching the B&R Automation community forum
This is a functional community MCP server with clearly named tools and documented schemas. All 5 tools have proper verb-noun naming (search_*, get_*, list_) and descriptions. Input schemas are visible and properly typed. However, several quality gaps prevent a higher score: (1) parameter descriptions are generic and lack actionable constraints (e.g., 'Search terms to look for' doesn't specify format, length, or example patterns); (2) output schemas are inferred from type annotations but not explicitly documented in docstrings; (3) error handling is minimal, no guidance on retries, user-fixable errors, or recovery paths; (4) no validation rules or constraints documented (e.g., max_posts range, period enum validation); (5) tool descriptions lack WHEN/WHY guidance, they state WHAT but not when to use one tool vs another. The server is well-structured for a read-only community forum tool, but falls short of A-grade quality due to incomplete parameter documentation and missing error recovery patterns.
Get the latest topics from the B&R Community forum.
Get the top/most popular topics from the B&R Community forum.
Get detailed information about a specific topic including its posts.
List all available categories in the B&R Community forum.
Search the B&R Community forum for topics and posts.
Enum constraints not formalized in JSON Schema. The get_top_topics 'period' parameter lists valid values in the description text ('daily', 'weekly', etc.) but doesn't register them as a JSON Schema enum. This relies on LLM text understanding instead of formal constraints, inviting hallucinated values and forcing validation errors at runtime.
Parameter descriptions lack actionable constraints and examples violate rubric. The search_community 'query' parameter includes example values in the description ('e.g., "mappView widget", "ACOPOS error"') which the rubric explicitly prohibits, LLMs latch onto examples and reuse them literally. Description should specify: min/max length, special syntax rules (e.g., how to use ##category or status:solved), character restrictions, or reference to Discourse search syntax docs.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 58 | 2026-07-28+ | v2 |
| 2026-03-09 | C | 62 | - | v1 |
Output schemas not documented in tool docstrings. While the code uses TypedDict models (SearchResults, TopicDetails, etc.) with proper structure, the tool descriptions do NOT document what fields are returned or their semantics. An LLM reading only the docstring cannot predict the output shape, forcing it to infer from trial. All tools should state: 'Returns: <StructName> with fields: <field1> (<type>), <field2> (<type>), ...'
No numeric bounds on integer parameters. The get_topic 'max_posts' parameter defaults to 10 but doesn't specify min/max. The get_latest_topics 'page' parameter lacks bounds (what page number is valid? does API cap at 1000?). Unbounded numbers let LLMs pass absurd values (page=999999, max_posts=10000000) causing API errors or timeouts. Every numeric parameter needs explicit min and max in the description.
No error handling guidance or recovery paths. None of the tool descriptions address what happens on failure: What if the search returns no results? What if topic_id doesn't exist? What if the API is rate-limited? Tools should include error classification (retryable, user-fixable, fatal) and guidance: 'If the topic is not found, try search_community to locate it by title.'
Tool descriptions lack WHEN/WHY guidance and mutual exclusivity hints. The descriptions state WHAT each tool does but not WHEN to use one over another. For example, get_latest_topics vs get_top_topics, when should the agent choose each? This forces the LLM to reason about intent without explicit guidance, increasing selection errors. Each description should include a 'Use when:' clause.
Optional parameters lack clear defaults and implications. The search_community 'category' parameter is optional but the description doesn't say what happens if omitted (searches all categories?). The 'solved_only' defaults to False but there's no context on the filter behavior. Optional parameters should document: default behavior if omitted, and why someone would use each variant.
Response limiting and pagination strategy not documented. The search_community 'posts' array and get_topic 'posts' array could be large. No mention of: max results returned, pagination token/cursor support, or how to retrieve next results.