Clew Study Assist MCP
An MCP server that shows your assistant the Clew learning graphs: topics with open prerequisites, notes and progress
- 213 ★ on the repository
- Last commit
- MIT
Minimal access
Why this level
- The server only reads the local SQLite database with your graphs and notes and never changes them
Install
- Works in
Manual install
claude mcp add "Clew Study Assist" /absolute/path/to/clew/.venv/bin/clew-study-assistInstall Clew from source first: git clone, cp .env.example .env and ./scripts/dev.sh. Replace the path with your clone's path.
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The server only reads the local SQLite database with your graphs and notes and never changes them
This is third-party code. Review the repository files before installing.
What it does
Clew builds a map of topics and prerequisites around what you want to learn and guides you through it with AI. The built-in MCP server `clew-study-assist` hands that map to another assistant such as Claude Code, Claude Desktop or Cursor over stdio. The assistant sees what you are studying, which topics are blocked by open prerequisites and what each note holds. There is no need to copy graph state into the chat. There are four tools and all of them only read. `list_graphs` returns every graph with progress stats. `get_current_learning_context` gives a summary of the active graph and the other graphs with activity. `get_node` returns one topic in full: description, resources, artifacts, neighbors and blocking prerequisites. `search_notes` searches titles, slugs and descriptions in one graph or across all of them. Graph edits stay in the Clew interface with its proposal review, and MCP changes nothing. The server reads the same SQLite database directly, so a running backend is not required. The server installs from source with the `./scripts/dev.sh` script, and the desktop installers do not configure external assistants.
Who it is for. Learners who study from a topic map in Clew and want Claude Code or Cursor to know what is already done and what blocks progress.
Good fit when
- You ask the assistant what to learn next and want an answer that accounts for open prerequisites
- You ask it to explain a topic from the graph and it needs the description, resources and neighboring topics
- You search your notes by keyword across several graphs at once
- You keep several graphs and the assistant should see progress in each
- You want another assistant to read the graph without any right to change it
Not a fit when
- Clew is not installed: the server reads that app's database and is useless without graphs
- You want to edit the graph from chat: edits do not go through MCP and are made in the Clew interface
- You expect a one-click install from the desktop installer: the installer creates no command on PATH
- Your notes live in another service: then you need that service's own MCP
Example request
Look at my Clew graphs and tell me which machine learning topics are blocked and what I should learn firstMCP
- Transport
- stdio
- Authentication
- not required
Environment variables
CLEW_DB_PATH- Path to the SQLite file holding the graphs, if you need a database other than the one in the clone.
Limitations
- The server installs from source: clone the repository and run `./scripts/dev.sh`. It needs Node 22.22.2 or newer, Python 3.11 or newer and an OS credential store.
- By default it reads the database in the clone directory. To see the desktop app's graphs, set the path to its database in an environment variable and find the location in the data section of `docs/DESKTOP.md`. The docs name this variable `CLEW_DB_PATH` in one place and `KG_DB_PATH` in another, so check the name on your version.
- Clew itself needs a ChatGPT Plus or Pro plan, and the 1.0.0 builds are not signed with Developer ID or Authenticode.
- The server only reads, and no action on the graph can be performed through MCP.
How to disable. Remove the `Clew Study Assist` entry from `mcpServers` in `claude_desktop_config.json` or `~/.claude.json`, or run `claude mcp remove "Clew Study Assist"`. For Cursor remove the server in the MCP settings. Restart the client. The graphs themselves are untouched.
FAQ
How do I connect Clew to Claude Code?
First install Clew from source, which produces the binary `.venv/bin/clew-study-assist`. Then run `claude mcp add "Clew Study Assist" /absolute/path/to/clew/.venv/bin/clew-study-assist`. The server appears in the MCP list after a restart.
Which tools does the server provide?
Four, all read-only: `list_graphs`, `get_current_learning_context`, `get_node` and `search_notes`. The assistant gets lists of graphs, a progress summary, topic content and note search results.
Can the assistant change my graph?
No. The server strictly reads the database. Edits go through the Clew interface on a propose-and-apply basis, and you approve each one.
How do I show the assistant the desktop app's graphs?
By default the server looks at the database in the clone. Set a variable in the server's `env` block with the path to the database from the app's data directory. The file location is given in the local data section of `docs/DESKTOP.md`.
How do I make the assistant always consult the graph?
Add one sentence to the client's custom instructions: before answering about what you study, call `get_current_learning_context`, and when a specific topic comes up use `search_notes` and `get_node`. The project documentation recommends this.
Discussion
No questions or reviews yet.
Related
MCP serversThe official Kibela MCP server searches, reads and edits notes, folders and comments of your wiki straight from an assistant chat
MCP server for the knowledge base Beever Atlas builds from Slack, Discord, Teams and Mattermost chats: cited answers, channel wikis, a graph
ChatCrystal: notes from agent conversations
ChatCrystal
A local knowledge base: imports Claude Code, Cursor and Codex chats into notes, semantic search and MCP memory
loci turns notes and agent memories into one knowledge base with hybrid search and answers cited down to a file section.
An MCP server for searching your Claude Code conversation history: past solutions, errors, plans and file change history.
Plugin and skill for Claude Code, Codex and Kilo that researches a topic on the web and builds a local learning board
README in short
The page presents Clew as a learning harness with an interactive knowledge graph in ChatGPT and in a desktop app. It separately covers app installation, data privacy, running for development and a map of directories. The MCP setup lives in `docs/MCP_SETUP.md`.