Local MCP server for PageIndex — interact with a self-hosted PageIndex installation without cloud APIs
Solid foundation with complete schemas and descriptions for all 8 tools. Tool names follow verb_noun convention (index_document, list_documents, search, etc.). All parameters have type definitions and descriptions. However, descriptions are often terse (many under 100 chars), parameter descriptions lack actionable constraints (e.g., no format hints for 'path' parameter), and output schemas are not documented. Error handling guidance is absent, tools don't explain what to do on failure. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. Missing idempotency guidance for write operations.
Return full metadata for one indexed document.
Return the PageIndex tree structure for a document, optionally limited by depth.
Check whether the local MCP server, PageIndex repo, Python environment, and workspace are configured correctly.
Add and index a local PDF or Markdown file using the local PageIndex installation. Returns document ID and tree path on success.
List documents in the local workspace registry.
Re-run local PageIndex generation for an existing document.
Output schemas not documented. Tools return results but LLMs cannot plan downstream calls without knowing response structure (e.g., what fields does get_document return? Does list_documents return total_count for pagination?)
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Destructive tools (remove_document) and write tools (index_document, reindex_document) should declare their impact so agents know which calls are safe to retry.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | B | 71 | <=2025-11-25 | v2 |
Remove a document from the local registry and optionally delete workspace artifacts.
Perform vectorless reasoning-based retrieval across locally indexed PageIndex documents using a local LLM.
Parameter descriptions lack actionable constraints. 'path' parameter says 'Absolute path to .pdf, .md, or .markdown file' but doesn't specify format validation, path traversal guards, or what happens if file doesn't exist. 'model' parameter is undescribed in several tools.
No error recovery guidance. Tools don't explain what to do on failure (e.g., if indexing fails, should agent retry? Call health check first? What are retryable vs fatal errors?). Agents cannot self-correct without this guidance.
Pagination support incomplete. list_documents accepts limit/offset but doesn't document whether total_count is returned, whether results are sorted, or what happens when offset exceeds total. Agents cannot reliably iterate without this clarity.