pantheon-ai/graphthulhu

Use and maintain the graphthulhu knowledge graph MCP to store investigations, architectural decisions, plans, and session context in a persistent Obsidian-backed vault; triggers automatically at session start, after analysis or investigations, when recording decisions, and when the user corrects stored knowledge.

Quality

99%

Does it follow best practices?

Impact

Pending

No eval scenarios have been run

Securityby

Passed

No known issues

graphthulhu Tool Catalogue

Name: pantheon-ai/graphthulhu
Rating: 0.99 (1 reviews)
Author: pantheon-ai

Complete reference for all graphthulhu MCP tools, grouped by purpose.

Status & Administration

Tool	Purpose
`health`	Check server status, version, backend type, read-only mode, page count. Use to verify the server is alive.
`analysis_health`	Diagnostic health check for the graph analysis subsystem.
`graph_overview`	High-level overview: total pages, blocks, links, most connected pages, orphan count, namespace breakdown. Call at session start.
`reload`	Reload the graph from disk. Use after external edits to the vault.

Reading Pages & Blocks

Tool	Purpose
`list_pages`	List pages with optional filters: namespace, property key, tag. Sorts by name, modified, or created.
`get_page`	Fetch a single page with all its blocks and properties.
`get_block`	Fetch a single block by UUID.
`get_links`	Get all pages that link to or from a given page.
`query_properties`	Query pages by property key/value pairs. Useful for finding pages by metadata.

Search & Discovery

Tool	Purpose
`search`	Full-text search across all blocks. Returns matching blocks with surrounding context (parent chain and siblings). Use `compact: true` to save tokens.
`find_by_tag`	Find all pages tagged with a specific tag.
`find_connections`	Discover how two pages are connected through the link graph. Returns direct links, shortest paths, and shared connections.
`traverse`	Walk the graph from a starting page up to a given depth. Useful for exploring a topic cluster.
`knowledge_gaps`	Identify pages that are referenced but don't exist yet, or pages with sparse content.
`list_orphans`	List pages with no incoming or outgoing links — isolated knowledge nodes.
`topic_clusters`	Group pages into topic clusters based on link density.

Writing Pages & Blocks

Tool	Purpose
`create_page`	Create a new page with optional properties and initial blocks. Use namespace/topic naming (e.g. `analysis/auth-flow`).
`upsert_blocks`	Batch-create blocks on a page. Supports nested children for hierarchies. Can append or prepend.
`append_blocks`	Append one or more blocks to an existing page.
`update_block`	Update the content or properties of a single block by UUID.
`delete_block`	Delete a single block by UUID.
`move_block`	Move a block to a different position or parent.
`bulk_update_properties`	Update properties on multiple pages at once.
`link_pages`	Create a bidirectional link between two pages.
`rename_page`	Rename a page and update all incoming links.
`delete_page`	Delete a page permanently.

Decisions

Decisions are tracked as #decision-tagged blocks on the page where context is richest.

Tool	Purpose
`decision_create`	Create a new `DECIDE` block with a question, options, and deadline.
`decision_check`	List all open decisions on a page.
`decision_resolve`	Mark a decision as resolved with the chosen option and rationale.
`decision_defer`	Mark a decision as deferred with a reason.