mcp-logseq MCP — Use Cases, Install & Live Demo

Why use it

Key features

16 tools for pages, blocks, namespaces, and Logseq DSL queries
Optional semantic vector search (vector_search, sync_vector_db)
Markdown-aware — create_page and update_page handle Logseq block formatting
DB-mode property support via LOGSEQ_DB_MODE
Tag exclusion filter (LOGSEQ_EXCLUDE_TAGS) to hide private pages from AI

Live Demo

What it looks like in practice

logseq.replay ▶ ready

0/0

Install

Pick your client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "logseq": {
      "command": "uvx",
      "args": [
        "mcp-logseq"
      ],
      "_inferred": true
    }
  }
}

Open Claude Desktop → Settings → Developer → Edit Config. Restart after saving.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "logseq": {
      "command": "uvx",
      "args": [
        "mcp-logseq"
      ],
      "_inferred": true
    }
  }
}

Cursor uses the same mcpServers schema as Claude Desktop. Project config wins over global.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "logseq": {
      "command": "uvx",
      "args": [
        "mcp-logseq"
      ],
      "_inferred": true
    }
  }
}

Click the MCP Servers icon in the Cline sidebar, then "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "logseq": {
      "command": "uvx",
      "args": [
        "mcp-logseq"
      ],
      "_inferred": true
    }
  }
}

Same shape as Claude Desktop. Restart Windsurf to pick up changes.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "logseq",
      "command": "uvx",
      "args": [
        "mcp-logseq"
      ]
    }
  ]
}

Continue uses an array of server objects rather than a map.

~/.config/zed/settings.json

{
  "context_servers": {
    "logseq": {
      "command": {
        "path": "uvx",
        "args": [
          "mcp-logseq"
        ]
      }
    }
  }
}

Add to context_servers. Zed hot-reloads on save.

claude mcp add logseq -- uvx mcp-logseq

One-liner. Verify with claude mcp list. Remove with claude mcp remove.

Use Cases

Real-world ways to use mcp-logseq

Draft today's daily journal from your week

👤 Logseq power users ⏱ ~10 min beginner

When to use: Friday reflection; you want a summary of the week's journal pages.

Prerequisites

Logseq with HTTP API enabled — Settings > Features > enable HTTP APIs; generate API token
LOGSEQ_API_TOKEN + LOGSEQ_API_URL env vars — Set in your MCP client config

Flow

Pull last week's journal pages

Get the content of this week's daily journal pages (Monday through today).✓ Copied

→ Page contents per day
Summarize and draft

Summarize themes and draft today's journal with a Week Review section.✓ Copied

→ Draft ready to paste
Write back

Update today's journal page, appending the Week Review section.✓ Copied

→ Page updated

Outcome: A closing-the-loop weekly reflection with minimal typing.

Pitfalls

update_page in append mode duplicates content on retry — Use replace mode for idempotent operations, or check existing content first
Sensitive tags leak into summaries — Set LOGSEQ_EXCLUDE_TAGS=private,#secret to hide those

Find pages by property for a task sweep

👤 Logseq users with task-style pages ⏱ ~15 min beginner

When to use: You tag project pages with status:: active and want a list.

Flow

Find by property

find_pages_by_property: status=active. Return names and updated-at.✓ Copied

→ Page list
Triage

For each, fetch the page and tell me the last 3 blocks — is there a visible next action?✓ Copied

→ Per-page status

Outcome: A sweep of active projects with clear next actions.

Semantic recall across your whole graph

👤 Researchers with a huge Logseq graph ⏱ ~10 min intermediate

When to use: Keyword search isn't finding a note you know you wrote.

Prerequisites

Vector DB synced — Run sync_vector_db once; re-run after big imports

Flow

Vector search

vector_search: 'what did I note about test pyramid vs honeycomb?'✓ Copied

→ Ranked matches
Pull full page

get_page_content on the top match.✓ Copied

→ Full content

Outcome: Finding notes by meaning, not exact words.

Turn a meeting page into tracked action items

👤 PMs and team leads ⏱ ~10 min beginner

When to use: You capture meetings in Logseq and want the actions separated.

Flow

Read meeting page

get_page_content for 'Meeting/2026-04-14/engineering-sync'.✓ Copied

→ Notes
Extract actions

Identify TODO-flavored lines. For each, create a child block under today's journal with assignee and due date.✓ Copied

→ Blocks created

Outcome: Actions lifted out of notes into a trackable list.

Combinations

Pair with other MCPs for X10 leverage

logseq + marm-systems

Short-lived recall in MARM; durable notes in Logseq

Move these MARM contextual logs into Logseq as pages under /Notes/2026-04.✓ Copied

logseq + filesystem

Export curated Logseq pages to Markdown files

Export all pages tagged #public to ./exports/public-notes/ as individual .md files.✓ Copied

Tools

What this MCP exposes

Tool	Inputs	When to call	Cost
list_pages	(none)	Graph overview	free (local API)
get_page_content	name: str	Read a page	free
create_page	name, content, properties?	New page	free (write)
update_page	name, content, mode: append\|replace	Edit an existing page	free (write)
delete_page	name: str	Destructive — confirm	free
search	query: str	Full-text search	free
query	dsl: str	Logseq DSL query	free
find_pages_by_property	key: str, value?: str	Filter by properties	free
get_page_backlinks	name: str	Who links to this note	free
insert_nested_block	parent_uuid, content, position: child\|sibling	Add structured content	free (write)
vector_search	query: str	Semantic recall	free (local embeddings)

Cost & Limits

What this costs to run

API quota: None — Logseq's local API is unbounded
Tokens per call: Small for list/search; pages can be long
Monetary: Free
Tip: For graphs with 10k+ pages, keep vector DB synced nightly rather than on every query.

Security

Permissions, secrets, blast radius

Credential storage: LOGSEQ_API_TOKEN in env var

Data egress: All local to Logseq; vector embeddings stay on your machine

Never grant: Exposing Logseq's HTTP API to your LAN without firewalling

Logseq's HTTP API has full graph write access — protect the token as a secret.
Use LOGSEQ_EXCLUDE_TAGS to redact private pages from AI visibility.

Troubleshooting

Common errors and fixes

Connection refused on localhost:12315

Start the HTTP API via the plug icon in Logseq. It's off by default.

Verify: curl http://localhost:12315

401 on API calls

Token missing or mismatched. Regenerate in Logseq Authorization tokens.

Verify: Check token in Logseq settings

vector_search empty

Run sync_vector_db first; it's not auto-synced.

Verify: vector_db_status

Alternatives

mcp-logseq vs others

Alternative	When to use it instead	Tradeoff
Obsidian MCP (community)	You use Obsidian instead	Different graph semantics
Notion MCP	You're on Notion	Paid SaaS; different graph model
filesystem MCP on the Logseq dir	You only need raw file access	No block-level operations, no DSL queries

More

Resources

📖 Read the official README on GitHub

🐙 Browse open issues

🔍 Browse all 400+ MCP servers and Skills