/setup

Deep Research Setup

Verify prerequisites for the deep-research plugin's multi-agent pipelines. All pipelines require Agent Teams; Pipeline D also requires a NotebookLM MCP server.

If $ARGUMENTS contains --check-only, report status without making changes.

---

1. Agent Teams (required)

echo "${CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS:-not_set}"

• If 1: ready.
• If not set: All pipelines will fail without this. Instruct the user to add to ~/.claude/settings.json:

  "env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" }
  

  Takes effect on next Claude Code restart.

---

2. Pipeline Availability

Check which pipelines are available by looking for their command files relative to this plugin:

PLUGIN_DIR="${CLAUDE_PLUGIN_ROOT}"
for cmd in web repo structured; do
  test -f "$PLUGIN_DIR/commands/$cmd.md" && echo "$cmd: available" || echo "$cmd: missing"
done

Also check for the NotebookLM sub-plugin:

test -d "$PLUGIN_DIR/notebooklm" && echo "notebooklm: available" || echo "notebooklm: not installed"

Report:

• Pipeline A (Internet Research) — /deep-research web
• Pipeline B (Repo Research) — /deep-research repo
• Pipeline C (Structured Research) — /deep-research structured
• Pipeline D (NotebookLM Media Research) — /notebooklm-research (requires notebooklm sub-plugin)

---

3. NotebookLM Setup (Pipeline D only)

Skip this section if the notebooklm sub-plugin is not installed.

3a. MCP server CLI

Check if notebooklm-mcp is available:

command -v notebooklm-mcp 2>/dev/null || npx --yes notebooklm-mcp --version 2>/dev/null || echo "not_found"

• If found: ready.
• If not found: Pipeline D requires the notebooklm-mcp-cli package. Install with:

  npm install -g notebooklm-mcp-cli
  

  Or see https://github.com/jacob-bd/notebooklm-mcp-cli

3b. Authentication

Note that NotebookLM requires Google account authentication. The user must run nlm login in their terminal (outside Claude Code) to authenticate. This handles OAuth flow and session cookie extraction.

If auth expires mid-session, the refresh_auth MCP tool or re-running nlm login will fix it.

3c. Enable/disable guidance

Note that the NotebookLM sub-plugin is kept disabled by default to reduce context load. The user should enable it in ~/.claude/settings.json before running Pipeline D research, and disable it after.

---

4. Status Report

## Deep Research Setup

| Check                       | Status |
|-----------------------------|--------|
| Agent Teams env var         | ... (REQUIRED) |
| Pipeline A (web)            | ... |
| Pipeline B (repo)           | ... |
| Pipeline C (structured)     | ... |
| Pipeline D (notebooklm)    | ... |
| NotebookLM MCP CLI         | ... (if Pipeline D) |
| NotebookLM auth             | run `nlm login` in terminal |

### Available commands

- `/deep-research web <topic>` — Internet research with iterative deepening
- `/deep-research repo <path>` — Repository assessment (add `--compare`, `--deeper`, `--deepest`)
- `/deep-research structured <spec>` — Schema-conforming batch research
- `/notebooklm-research <topic>` — Media research (YouTube, podcasts, audio)

If Agent Teams is not set, make this prominent — nothing will work without it.

End with: "Run /deep-research web 'test topic' to verify the setup works."