Skip to main content
Glama
Sign In
ultrathink-mcp

ultrathink-mcp

by Fifty-Five-and-Five
OverviewSchemaSourceRelated ServersDiscussionsAdmin
JavaScript
Hybrid

Ultrathink MCP Server

Save and search your Ultrathink knowledge base from any MCP-compatible client: Claude Desktop, Claude Code, Cursor, Zed and more.

Ultrathink is a second-brain app that captures notes, links, screenshots, AI conversations, PDFs and more into a single AI-searchable knowledge base.

New: the hosted Ultrathink MCP server

Ultrathink now runs its own MCP server online, so there is nothing to install. It has more tools than this package: search, read an entry in full, related entries, recent changes, everything about a person or a topic, lists of your topics and people, and adding and updating entries.

Connect Claude Code with one command, using a device key with API access:

claude mcp add --transport http ultrathink https://us-central1-ultrathink-50f7f.cloudfunctions.net/mcp --header "Authorization: Bearer YOUR_DEVICE_KEY"

Any MCP client that supports remote servers over HTTP with a header can use the same address. Setup, tools and troubleshooting are in the MCP server guide.

The hosted server is the recommended way to connect. This package keeps working, with the two tools below, for clients that only run local servers.

Related MCP server: open-brain-mcp

This package

It gives your AI assistant two tools:

  • ultrathink_save: store a note, link, idea or snippet in your knowledge base

  • ultrathink_search: natural-language search across everything you've captured

Install

npm install -g ultrathink-mcp

Or use directly with npx ultrathink-mcp (no install needed).

Configuration

You need an Ultrathink device key. Get one at tryultrathink.com.

Set as environment variables:

export ULTRATHINK_DEVICE_KEY="your-device-key-here"

Optional:

export ULTRATHINK_API_BASE="https://us-central1-ultrathink-50f7f.cloudfunctions.net"  # default
export ULTRATHINK_SOURCE="claude-code"  # tag entries with a source label

Use with Claude Code

Add to your ~/.claude.json mcpServers:

{
  "mcpServers": {
    "ultrathink": {
      "command": "npx",
      "args": ["-y", "ultrathink-mcp"],
      "env": {
        "ULTRATHINK_DEVICE_KEY": "your-device-key-here"
      }
    }
  }
}

Use with Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "ultrathink": {
      "command": "npx",
      "args": ["-y", "ultrathink-mcp"],
      "env": {
        "ULTRATHINK_DEVICE_KEY": "your-device-key-here"
      }
    }
  }
}

Tools

ultrathink_save

Save an entry to your Ultrathink knowledge base.

Inputs:

  • title (required): title of the entry

  • notes: main content for most entries

  • content: full content for longer entries

  • type: one of note, link, idea, snippet, para, chatgpt, claude, perplexity, notion, long-note, profile. Default: note

  • url: required when type is link

  • visibility: private (default) or org

ultrathink_search

Natural-language search across your Ultrathink knowledge base. Returns an AI-generated answer based on stored entries.

Inputs:

  • query (required): natural-language search query

  • entityFilter: optional, one of project, task, knowledge

Licence

MIT

Available Tools

2 tools
ultrathink_saveB

Save an entry to Ultrathink knowledge base. Use this to store notes, ideas, links, snippets, and other knowledge.

NameRequiredDescriptionDefault
urlNoURL (required for link type)
typeNoType of entry. Default: note
notesNoNotes or description (main content for most entries)
titleYesTitle of the entry
contentNoFull content (for longer entries)
visibilityNoVisibility: private (default) or org (shared with organization)

TDQS

B3.2/5.0
Behavior2/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

No annotations are provided, so the description carries the full burden and only conveys that this is a write. It says nothing about what is returned (an ID?), whether duplicate titles overwrite or create a second entry, permission or org-sharing behavior behind visibility, or what happens if type=link without a url.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Two short sentences with the action leading and the usage examples second – appropriately sized and front-loaded. The second sentence borders on generic ('other knowledge') but does not waste much space.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

The rich input schema compensates for much of the missing detail, and no output schema exists so return values needn't be spelled out. However, for a 6-parameter mutation tool with zero annotations, the description leaves an agent without the create-vs-update semantics or failure behavior it would need to call this confidently.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is 100%, so both enum values and each field's meaning (url required for link, visibility private/org, type default note) are already documented. The description adds nothing beyond restating a subset of the content types, so the baseline 3 applies.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb+resource ('Save an entry to Ultrathink knowledge base') and illustrates the kinds of content it accepts. It separates itself from the only sibling, ultrathink_search, by implication (save vs. search), though it never names the alternative.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The second sentence gives examples of when to reach for it (notes, ideas, links, snippets), which implies the trigger condition. There is no explicit when-not guidance, no mention of updating vs. creating, and no reference to ultrathink_search as the retrieval counterpart.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 2 tool updatesv1.0.1
    • First observedultrathink_save
    • First observedultrathink_search
Deploy Server
Context All Your Agents Can ShareAgents share what they learn, hand off work, and hold it when you switch models. Start with 5 GB free.fulcradynamics.comAd

TDQS

A3.5/5.0

Scored across 2 tools

Disambiguation5/5

The two tools perform clearly opposite, non-overlapping operations: write (save) and read (search). An agent can instantly choose between storing knowledge and retrieving it, with no ambiguity.

Naming Consistency5/5

Both tools follow an identical pattern: the `ultrathink_` namespace prefix plus a concise verb (save, search). Fully predictable and consistent.

Tool Count3/5

Two tools is on the thin side for a knowledge base server; a minimal capture-and-retrieve surface works, but there is little else an agent can do. It sits at the borderline of under-scoped.

Completeness3/5

Core store-and-query flow is covered, but obvious knowledge-management operations are absent: no update, delete, list/recent entries, or get-by-id. Agents cannot correct or manage stored entries, only add and search.

Maintenance

ActivityMaintained
ResponsivenessNo issues

Resources

Related MCP Connectors

Related MCP Servers

  • open-brain-mcp

    uwek
    A
    license
    Not graded
    quality
    D
    maintenance
    An MCP server for storing and retrieving thoughts, notes, and ideas with semantic vector search, enabling integration with Claude Desktop and other MCP clients.
    MIT

  • memory-mcp

    AyubMoh1
    A
    license
    Not graded
    quality
    D
    maintenance
    Persistent memory MCP server for Claude Code. Store, search, and retrieve memories across sessions using hybrid vector + keyword search powered by SQLite, FTS5, and sqlite-vec.
    8 npm
    1
    MIT