graphthulhu

Knowledge graph MCP server for Logseq and Obsidian. 37 tools for reading, writing, searching, and analyzing your second brain.

Security Scan

The SKILL.md and required-binary declaration mostly match the described MCP server purpose, but there are inconsistencies and modest install/credential risks you should review before installing.

Purpose & Capability

The declared required binary (graphthulhu) and the SKILL.md both describe an MCP server for Logseq and Obsidian, which is coherent. However registry metadata lists no required env vars or source/homepage, while SKILL.md points to a GitHub project and documents LOGSEQ_API_URL and LOGSEQ_API_TOKEN usage — a metadata mismatch that should be clarified.

Instruction Scope

Instructions are scoped to running a local graphthulhu binary and configuring MCP entries for Obsidian (vault path) or Logseq (HTTP API URL + token). That aligns with the stated functionality, but the skill explicitly requires full read/write access to your vault (create/update/delete/rename blocks/pages), which is high-privilege — expected for this purpose but worth explicit consent and backups.

Install Mechanism

No formal install spec in the registry; SKILL.md tells users to download a binary from GitHub Releases or run `go install github.com/skridlevsky/graphthulhu@latest`. Both options execute third-party code on the machine. Downloading arbitrary binaries or building from a remote repo carries moderate risk and should be audited (verify releases, checksums, and source code) before running.

Credentials

Registry declares no required env vars, but the Logseq example uses LOGSEQ_API_URL and LOGSEQ_API_TOKEN (credentials that grant access to local Logseq content). Requiring a token for Logseq is proportionate to the feature, but the omission from metadata is a discrepancy; confirm required env vars and limit token scope where possible.

Persistence & Privilege

The skill is not marked always:true and does not request persistent platform-wide privileges. It will invoke a local binary you install; autonomous agent invocation is allowed by default (normal), so be aware an agent could call this binary if the skill is enabled.

What to consider before installing

This skill appears to be what it says (an MCP server for Logseq/Obsidian) but take these precautions before installing: 1) Confirm the GitHub repository and inspect the source/release assets (checksums, signatures) before running the binary or using `go install`. 2) Recognize the tool gets full read/write access to your vault — back up your notes and restrict the Logseq token scope. 3) Ask the publisher or registry maintainer to update metadata to list the LOGSEQ_API_URL/LOGSEQ_API_TOKEN requirement and the GitHub source. 4) Prefer running the binary in a controlled environment (sandbox or dedicated machine) until you trust it. If you want higher confidence, request a signed release, checksum, or an install spec from a well-known package host.

MCP server that gives you full access to a Logseq or Obsidian knowledge graph. 37 tools across 9 categories: navigate, search, analyze, write, decisions, journals, flashcards, whiteboards, and health.

Install

Download the binary for your platform from GitHub Releases and put it on your PATH.

Or: go install github.com/skridlevsky/graphthulhu@latest

Configure

Obsidian

Add to your MCP settings:

{
  "mcpServers": {
    "graphthulhu": {
      "command": "graphthulhu",
      "args": ["--backend", "obsidian", "--vault", "/path/to/your/vault"]
    }
  }
}

No plugins required. Reads .md files directly. Full read-write support.

Logseq

Enable the HTTP API server in Logseq (Settings > Features > HTTP APIs server), start it, and create a token.

{
  "mcpServers": {
    "graphthulhu": {
      "command": "graphthulhu",
      "env": {
        "LOGSEQ_API_URL": "http://127.0.0.1:12315",
        "LOGSEQ_API_TOKEN": "your-token-here"
      }
    }
  }
}

What you can do

• Navigate — get pages with full block trees, traverse the link graph, list pages by namespace/tag/property
• Search — full-text search with context, property queries, tag hierarchy search, raw Datalog (Logseq)
• Analyze — graph overview, find connections between pages, detect knowledge gaps and orphans, discover topic clusters
• Write — create pages, append/upsert blocks with nested children, update/delete/move blocks, link pages bidirectionally, rename pages with link updates, bulk update properties
• Decisions — create decisions with deadlines, check status, resolve or defer with tracking
• Journals — read entries by date range, search within journals
• Flashcards — SRS stats, due cards, create new cards (Logseq)
• Whiteboards — list and inspect spatial canvases (Logseq)

Links

• GitHub
• Full tool reference