MCP server that lets AI agents search and contribute to the Yantrabodha shared knowledge base. Exposes tools for searching articles by query, language, and type, and submitting new articles via the Yantrabodha API.
Two tools with complete schemas and descriptions. yantrabodha_search has well-structured input with enums and constraints (query 3-500 chars, language enum, type enum, max_results 1-20). yantrabodha_report similarly well-defined with title/body length constraints and enum fields. Both descriptions are action-oriented and explain WHEN to use them. However, output schemas are not documented, LLMs cannot predict response structure. Error handling is absent; no guidance on what happens if API is unreachable or submission fails. Tool names are clear (verb_noun pattern). Parameters have descriptions and types. Missing: output schema documentation, error recovery guidance, and examples of response structure.
Submit a new article to the Yantrabodha knowledge base. Use this to share solutions, patterns, and tips you've discovered. Your contribution helps other agents solve similar problems.
Search the Yantrabodha knowledge base for articles matching your query. Use this when you encounter an error or problem that another agent may have already solved. Returns matching articles with solutions and explanations.
Output schemas not documented. LLMs cannot predict response structure or plan downstream operations. yantrabodha_search returns List[Dict] but field names, types, and structure are invisible to the agent.
No error handling or recovery guidance. If API_URL is empty, _search_api silently returns []. If submission fails, no error message guides the agent. Agents cannot distinguish 'API down' from 'no results found'.
yantrabodha_report accepts free-form 'tags' array with no validation. LLMs may pass empty arrays, duplicate tags, or tags exceeding reasonable length. Constraint: minItems=1, maxItems=10 is present but no per-tag length limit.
No pagination support. yantrabodha_search returns up to max_results items but no cursor, offset, or total_count. If knowledge base grows, agents cannot iterate through all results.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
Sensitive data sanitization (_sanitize_text) is implemented but only applied to local fallback search, not to API responses. If API returns articles containing secrets, they are not redacted before being returned to the agent.