Installing in Claude Code

This page installs PRIMA Scholar 2.2.1 as a Claude Code plugin, which is the recommended route. The plugin stays installed between sessions and works in every folder you open in Claude Code. A second route, which copies PRIMA Scholar into a single workspace, is described further down.

Before You Start

You need Node.js 18 or later, npm and git on your computer, as well as Claude Code itself. The Requirements page covers the Claude Code side of this list.

Install From the Plugin Marketplace

1. In Claude Code, add the PRIMA Scholar marketplace with this command.

   /plugin marketplace add larrygmaguire-hash/prima-scholar

2. Install the plugin from the marketplace you have just added with this command.

   /plugin install prima-scholar@prima-scholar

   On Claude Code 2.1.275 or later, one command does both steps.

   /plugin install prima-scholar --marketplace larrygmaguire-hash/prima-scholar

3. Open a terminal and build the search server inside the installed copy.

   cd ~/.claude/plugins/cache/prima-scholar/prima-scholar/2.2.1/mcp-servers/prima-scholar-search-mcp && npm install && npm run build

4. Build the library server in the same way, from its own folder in the installed copy.

   cd ~/.claude/plugins/cache/prima-scholar/prima-scholar/2.2.1/mcp-servers/prima-scholar-library-mcp && npm install && npm run build

5. Quit Claude Code and start it again so it loads the plugin and both servers.

The build steps are needed because the plugin download does not include the servers’ built files. Claude Code keeps the installed copy in ~/.claude/plugins/cache/prima-scholar/prima-scholar/ with a folder for each version, so the build always happens inside the version folder.

Check That It Works

Type /mcp in Claude Code and look for the two PRIMA Scholar servers, whose names end in prima-scholar-search and prima-scholar-library. Then ask for a citation with a DOI taken from the product’s own examples.

/prima-scholar:cite 10.1207/S15327965PLI1104_01

Claude looks the DOI up and returns an APA7 citation for the paper. If either server is missing from the list, see Troubleshooting.

Command Names in a Plugin Install

Claude Code puts the plugin name in front of every command and skill that a plugin provides. The four commands therefore appear as /prima-scholar:scholar, /prima-scholar:cite, /prima-scholar:library and /prima-scholar:literature-watch.

Add API Keys and Contact Emails

PRIMA Scholar starts with no keys set, and nine of the ten databases work straight away. On the plugin route the servers read their settings from the environment that Claude Code starts in. Export each setting in your shell before you start Claude Code, as API Keys and Configuration explains.

Update PRIMA Scholar

1. Refresh the PRIMA Scholar marketplace listing in Claude Code with this command.

   /plugin marketplace update prima-scholar

2. Update the installed plugin to the latest version with this command.

   /plugin update prima-scholar@prima-scholar

3. Build both servers again inside the new version folder, using the commands from the install steps with the new version number in the path.

4. Quit Claude Code and start it again so it loads the new version of both servers.

Remove PRIMA Scholar

1. Uninstall the PRIMA Scholar plugin from Claude Code with this command.

   /plugin uninstall prima-scholar@prima-scholar

2. If you no longer want the marketplace listed, remove it with this command.

   /plugin marketplace remove prima-scholar

Nothing was copied into your workspace on this route, so there is nothing else to remove there.

Caution

Uninstalling the plugin leaves your research library in place. The library is the file ~/.research-library/library.db unless you chose another location, together with any -wal and -shm files beside it. Deleting those files deletes every record, tag and collection in the library.

Install Into One Workspace With the Installer

PRIMA Scholar also ships an older installer, install.sh, which copies the commands, skills, agents and servers into one workspace’s .claude/ folder. It then registers the two servers in ~/.claude.json. This route suits you only if you want PRIMA Scholar to live inside one workspace.

The installer needs bash and python3, so it runs on macOS and Linux but not on Windows outside a bash environment. The workspace must already have a .claude/ folder.

You can give Claude Code the repository link and ask it to install PRIMA Scholar with the installer, because the README carries step-by-step instructions written for Claude. To do it yourself, follow these steps.

1. Clone the repository to ~/Developer/prima-scholar or another folder that sits outside your workspace.

   git clone https://github.com/larrygmaguire-hash/prima-scholar.git ~/Developer/prima-scholar

2. Build the search server inside your clone with this terminal command.

   cd ~/Developer/prima-scholar/mcp-servers/prima-scholar-search-mcp && npm install && npm run build

3. Build the library server inside your clone with this terminal command.

   cd ~/Developer/prima-scholar/mcp-servers/prima-scholar-library-mcp && npm install && npm run build

4. Run the installer and give it the path to your workspace.

   ~/Developer/prima-scholar/install.sh /path/to/workspace

5. Quit Claude Code and start it again so it loads the new commands and servers.

Caution

Never clone PRIMA Scholar inside your workspace folder. Git then treats it as a repository nested inside your own, which causes warnings and broken commits. The installer refuses to run from inside the workspace and tells you where to move the clone.

What the Installer Changes

Location	Change
.claude/commands/	Copies cite.md, library.md, literature-watch.md and scholar.md
.claude/skills/	Copies the four skill folders, replacing any folder with the same name
.claude/agents/	Copies the three agent files
.claude/mcp-servers/	Copies the built servers into prima-scholar-search/ and prima-scholar-library/, each with a start.sh script
~/.claude.json	Registers both servers at user scope, keeping any settings already in their env blocks
.gitignore	Adds the two server folders, so you do not commit them

If a server has not been built, the installer skips it, prints the build command and does not register it, so build both servers first. If the workspace has no .gitignore file, the installer prints the two lines for you to add yourself.

If ~/.claude.json does not exist, the installer skips the registration and you add the servers yourself. On this route you set API keys in the env block of each server entry in ~/.claude.json, then restart Claude Code.

Updating an Installer Install

1. Pull the latest version of PRIMA Scholar into your clone with this command.

   cd ~/Developer/prima-scholar && git pull

2. Rebuild both servers with the commands from steps 2 and 3 above.

3. In your workspace, delete the two server folders that the last install created, .claude/mcp-servers/prima-scholar-search and .claude/mcp-servers/prima-scholar-library.

4. Run the installer again and give it the path to your workspace.

5. Quit Claude Code and start it again so it loads the updated servers.

Step 3 matters because the installer copies each server’s build and node_modules folders with a plain copy. When those folders already exist, the new files land inside them and the old server code keeps running. Your API keys live in ~/.claude.json, and your library lives elsewhere, so deleting the server folders loses neither.

Caution

The installer replaces every command, skill and agent that has a name PRIMA Scholar ships. Any edits you made to those files are lost on the next run. To keep a customised version, copy the file or folder to a new name first, for example my-researching-topics, and edit the copy.

Removing an Installer Install

1. From your workspace folder, remove the commands, skills and agents the installer copied.

   rm .claude/commands/cite.md .claude/commands/library.md .claude/commands/literature-watch.md .claude/commands/scholar.md
   rm -r .claude/skills/literature-watch .claude/skills/managing-research-library .claude/skills/researching-topics .claude/skills/writing-with-citations
   rm .claude/agents/form-completion-agent.md .claude/agents/literature-watch-agent.md .claude/agents/research-agent.md

2. Remove the two server copies from your workspace with this command.

   rm -r .claude/mcp-servers/prima-scholar-search .claude/mcp-servers/prima-scholar-library

3. Remove the two server registrations, which also removes any API keys stored in them.

   claude mcp remove prima-scholar-search -s user
   claude mcp remove prima-scholar-library -s user

4. Delete the two lines that the installer added to your workspace .gitignore file.

   .claude/mcp-servers/prima-scholar-search/
   .claude/mcp-servers/prima-scholar-library/

5. Quit Claude Code and start it again so the removed commands and servers disappear.

Renamed copies such as my-researching-topics are yours and are not in this list. The research library and any literature watch files in .claude/literature-watch/ stay in place, and you can delete the clone at ~/Developer/prima-scholar when you no longer need it.

For Developers

To work on PRIMA Scholar itself, clone the repository, build both servers and start Claude Code with claude --plugin-dir ~/Developer/prima-scholar. This option loads the plugin for that session only, and nothing is copied into your workspace or the plugin cache.