Server has 20 tools with mostly present descriptions and schemas, but quality is inconsistent. Strong points: create_page has excellent, detailed description (250+ chars) with examples; vector_search has well-structured schema with enums and constraints. Weak points: 12 tools have minimal descriptions (under 50 chars: list_pages, get_page_content, delete_page, delete_block, update_block, get_block, search, query, find_pages_by_property, get_pages_from_namespace, get_pages_tree_from_namespace, rename_page, get_page_backlinks, insert_nested_block, set_block_properties, vector_db_status). Most parameter descriptions are absent or generic. No output schemas documented. Error handling guidance is minimal. Naming is consistent (verb_noun pattern) and clear.
Create a new page in Logseq with properly structured blocks. Fails if a page with the same title already exists (use update_page to modify existing pages). This makes retries safe: re-sending a create_page that timed out will not create numbered duplicates like "Page(1)". Markdown content is automatically parsed into Logseq's block hierarchy: - Headings (# ## ###) create nested sections - Lists (- or 1.) become proper block trees - Code blocks are preserved as single blocks - YAML frontmatter (---) becomes page properties Example content: ``` --- tags: [project, active] priority: high --- # Project Title Introduction paragraph. ## Tasks - Task 1 - Subtask A - Task 2 ```
Delete a specific block by UUID
Delete a page from Logseq
Find pages that have specific property values
Get a specific block by UUID
12 of 20 tools have minimal descriptions (<50 chars). LLMs cannot determine when to select these tools. Examples: list_pages, get_page_content, delete_page, search, query all lack context for selection.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. Agents cannot chain tools effectively without knowing what fields are returned.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 37 | <=2025-11-25 | v2 |
Get all pages that link to a specific page
Get the full content of a page including blocks and properties
Get all pages from a specific namespace
Get pages from a namespace as a hierarchical tree
Insert a block with nested children at a specific location
List all pages in the Logseq graph
Query Logseq using DataScript queries
Rename a page in Logseq
Search for content across Logseq pages and blocks
Set properties on a specific block
Run incremental sync of the vector database against current Logseq graph files. Only changed files are re-embedded. Use rebuild=true to drop and re-index everything.
Update the content of a specific block
Update an existing page in Logseq with new content or properties
Get the status of the vector database including sync state and staleness information
Semantic search over Logseq notes using hybrid search (vector similarity + full-text, combined by default). Use this for natural language queries about topics, concepts, or meaning — not for exact title lookups. Results include a relevance label and score (lower score = more relevant). Weak match results (score > 0.80) may be tangential; use judgment when presenting them to the user.
Parameter descriptions are missing or generic across most tools. update_page, delete_page, search, query lack any parameter documentation. LLMs cannot infer what values are valid or required.
Destructive tools (delete_page, delete_block) lack confirmation or dry-run support. No error guidance for recovery. Agents can permanently delete data without safeguards.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) present. Clients cannot distinguish safe reads from destructive writes at the protocol level.