Claude Code

Claude Code (Anthropic's CLI for terminal-based coding sessions) talks to MCP servers via the same JSON-config pattern as Claude Desktop. Setup is one command.

1. Get a Papermark token

Mint one in the dashboard at Settings → API Tokens. Read-only scopes are a good default; add write scopes if you want the agent to create links, datarooms, or upload documents.

2. Register the MCP server

claude mcp add papermark \
  --command npx \
  --args "-y" "@papermark/mcp-server" \
  --env PAPERMARK_TOKEN=pm_live_…

That's the entire setup. The next time you start a session, Claude Code spawns the MCP server and registers all 18 tools.

You can also edit ~/.config/claude-code/config.json (or wherever Claude Code stores its config — claude config path will tell you) directly:

{
  "mcpServers": {
    "papermark": {
      "command": "npx",
      "args": ["-y", "@papermark/mcp-server"],
      "env": {
        "PAPERMARK_TOKEN": "pm_live_…"
      }
    }
  }
}

3. Verify

Start a Claude Code session and run:

/mcp

You should see papermark in the list of connected servers, with its tool count. Then ask:

List my Papermark documents.

When to reach for it

Claude Code is at its best when MCP usage interleaves with shell commands and file edits — exactly what Papermark workflows look like:

• "Pull the latest investor memo from Papermark, save it to ~/Documents/, and open it." (get_document → file write → open shell command)
• "Find every link to the Q4 deck created in the last 30 days, dump the view counts to a CSV." (list_links → list_link_views → shell csv formatting)
• "Sync our Acme Drive folder into the matching Papermark dataroom." (filesystem walk → upload_document per file → list_dataroom_documents to verify)

Because Claude Code can read and write files, the upload_document tool — stdio-only — is genuinely useful here in a way it isn't from a browser-hosted client.

Per-project setup

If you want different tokens per project (e.g., production token in your prod-ops repo, test token in your sandbox repo), put the config in the project's local .claude/config.json instead of the global config file. Claude Code merges the two, with project-level overriding global.

Troubleshooting

Same playbook as Claude Desktop:

• Empty tool list → JSON syntax error
• No API token configured → missing PAPERMARK_TOKEN in env
• 401 → token revoked or wrong environment
• 403 on a write tool → mint a token with the missing scope

claude mcp logs papermark shows the server's stderr if Claude Code exposes that command in your version; otherwise the server's logs are wherever Claude Code stores MCP server logs (check claude --help for the path).