NLM Expert Research MCP — Frontier
A private, single-owner remote MCP that exposes a read-oriented NotebookLM research surface: grounded research, notebook/source discovery, evidence retrieval and comparison, plus auth and health diagnostics.
Grounded research —
researchqueries exactly one notebook and returns a source-locked evidence capsule (claims, citations, conflicts, gaps, quality status), tunable viamode(AUTO/INSTANT/STANDARD/DEEP/VERIFY/ULTRA/FORENSIC),freshness, andinternet_policy(LOCKED/AUTO/VERIFY/AUGMENT/ULTRA).Notebook discovery —
list_notebooksenumerates live notebooks with an explicit count and flagscatalog_stalewhen live discovery fails;notebook_resolveresolves by title, alias, ID or URL (cache never authoritative).Source scope —
list_sourcesreturns authoritative remote source IDs/titles via the RPC-only metadata path; note source-title scoping is prompt-scoped, not a strict transport-level filter.Catalog refresh —
notebook_refreshforces a live refresh and detects created, renamed, deleted and newly added items.Evidence reuse —
get_evidenceretrieves a single evidence item by ID from a prior research session.Cross-notebook comparison —
compare_notebookscompares two notebooks while preserving provenance per notebook; never invoked implicitly.Auth bootstrap —
setup_authopens a visible Chrome window for secure Google login (force=trueif cookies look valid but Google redirects to sign-in).Diagnostics —
healthchecks Frontier, persistent NotebookLM auth and upstream transport readiness without exposing credentials;doctorruns local runtime/adapter/registry/stdio-safe MCP checks.Boundaries — one authorized owner, read-oriented, notebook/source bodies returned only to the authenticated caller and never persisted or logged, and fail-closed on upstream RPC/UI drift;
VERIFYis source-grounded model evaluation, not deterministic proof.
Integrates with Google NotebookLM, allowing users to treat each notebook as a closed corpus, preserve evidence provenance, and retrieve compact evidence capsules. Provides tools for research, notebook resolution and refresh, evidence retrieval, and comparing notebooks.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@NLM Expert Research MCP — Frontier compare evidence on aspirin vs clopidogrel across my notebooks"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
NotebookLM Expert Research MCP for Perplexity
Independent remote MCP for one authorized owner. It exposes a compact, read-oriented NotebookLM surface over authenticated Streamable HTTP and keeps its browser/auth profile, metadata, token, runtime and service separate from the Claude Desktop product.
Product boundary
PRODUCT_MODE=PRIVATE_SINGLE_OWNEROne authorized Google identity and one owner-controlled remote MCP client.
No multi-user/SaaS account model.
Notebook/source bodies are returned only to the authenticated caller and are not persisted or logged.
The NotebookLM adapter is an explicit
HIGH_EXTERNAL_UI_DEPENDENCY; v1 prefers the upstream internal RPC path and fails closed when its contract or UI fallback drifts.
Related MCP server: OneClickLM
Local verification
npm.cmd install
npm.cmd run certify:localThe automated suite uses TEST_ONLY_FAKE_ADAPTER only under NODE_ENV=test. These checks certify server behavior, isolation and MCP protocol mechanics; they are not evidence of live NotebookLM authentication or research.
Staging runtime
Copy .env.example to a root-readable staging environment file, generate a unique random token of at least 32 characters, and keep the service bound to 127.0.0.1. Use the artifacts in deployment/ for the dedicated Unix user, systemd service and nginx HTTPS reverse proxy.
The one-time legitimate Google ceremony uses the loopback-only noVNC scripts documented in docs/AUTH_BOOTSTRAP.md. Its auth command is:
sudo -u nlm-perplexity-stage /opt/nlm-expert-research-mcp-perplexity-staging/deployment/auth-bootstrap-staging.shIt must run only through a secure interactive display method described in docs/AUTH_BOOTSTRAP.md. Never paste passwords, 2FA codes or cookies into chat.
Tool policy
nlm_consultar: ordinary targeted question.nlm_investigarSTANDARD: bounded synthesis.nlm_investigarDEEP: bounded primary plus counter-evidence/gap pass when the measured remaining budget permits it.nlm_verificar_respuesta: NotebookLM source-grounded model evaluation, not deterministic proof.Source title scoping in v1 is disclosed as prompt-scoped, not a strict transport-level source filter.
See docs/OPERATIONS.md, docs/AUTH_BOOTSTRAP.md, docs/UI_DRIFT_RUNBOOK.md, and docs/PERPLEXITY_ACCEPTANCE.md.
Available Tools
10 toolscompare_notebooksC
Explicitly compare two notebooks while preserving provenance by notebook. Never invoked implicitly.
Inputs: query (required), notebook_a (required), notebook_b (required).
doctorB
Run local diagnostics for runtime, adapter, registry and stdio-safe MCP operation.
Inputs: no named parameters.
get_evidenceA
Retrieve one evidence item by evidence ID from a previous research session.
Inputs: evidence_id (required).
healthA
Check Frontier, persistent NotebookLM authentication, and upstream transport readiness without exposing credentials.
Inputs: no named parameters.
list_notebooksA
List the authenticated account’s live NotebookLM notebooks. Use this when the user asks what expert notebooks are available; the result includes an explicit count and cached results are marked catalog_stale when live discovery is unavailable.
Inputs: freshness.
list_sourcesA
List authoritative remote source IDs and titles for one NotebookLM notebook using the RPC-only metadata path. Use before discussing corpus scope or freshness.
Inputs: notebook (required).
notebook_refreshB
Force refresh of live notebooks and sources, detecting created, renamed, deleted and newly added items.
Inputs: notebook_id.
notebook_resolveA
Resolve a notebook by title, alias, ID or URL. Cache misses trigger a live refresh; cache is never authoritative.
Inputs: notebook (required), freshness.
researchA
Research a question against exactly one NotebookLM notebook. Returns a compact source-locked evidence capsule with claims, citations, conflicts, gaps and quality status. Use this for substantive notebook questions.
Inputs: query (required), notebook (required), mode, freshness, +1 more.
setup_authA
Open a visible Chrome window for secure Google login. Set force=true when cookies look valid locally but Google redirects to sign-in.
Inputs: force.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 10 tools
Tools have mostly distinct purposes: setup_auth is for login, health for status, doctor for local diagnostics, research for single-notebook queries, compare_notebooks for two-notebook comparisons, and so on. Some potential confusion exists between list_notebooks and notebook_resolve (listing vs resolving one) and doctor vs health, but descriptions clarify the boundaries well.
Naming follows a mostly verb_noun pattern (list_notebooks, get_evidence, compare_notebooks) but deviates with single-word names (doctor, health, research) and reversed object_verb forms (notebook_resolve, notebook_refresh). This mixed convention is readable but not fully consistent.
10 tools is well-scoped for an expert research MCP, covering authentication, diagnostics, health, notebook discovery, source listing, research, evidence retrieval, and comparison. Each tool earns its place without redundancy or bloat.
The surface covers the full research lifecycle: setup auth, check health, list/resolve/refresh notebooks, list sources, run research, retrieve evidence, and compare notebooks. Minor gaps include lack of a tool to browse all evidence from a session or a single tool to search across multiple notebooks, but these are not core to the stated purpose.
Maintenance
Related MCP Connectors
Augments MCP Server - A comprehensive framework documentation provider for Claude Code
The Remote MCP server acts as a standardized bridge between LLM applications (like Claude, ChatGPT, and Cursor) and external services, enabling AI agents to access external tools and resources. Its primary capability is providing a centralized search tool to discover other MCP servers and their respective tools. Unlike local implementations, it runs remotely with OAuth authentication and permission controls for security.
MCP server giving Claude AI access to 22+ NYC public-record databases for real estate due diligence
Shared memory MCP server for Claude Code, Codex, Gemini CLI and Cursor. Cited, editable.
Related MCP Servers
- AlicenseAqualityFmaintenanceAn enhanced MCP server that enables users to interact with NotebookLM while ensuring strict source fidelity and citation accuracy through client-side prompt structuring. It guides Claude to transform user queries into structured prompts that prevent the addition of external knowledge during document analysis.1642MIT
- AlicenseAqualityDmaintenanceA robust MCP server for NotebookLM that provides auto-healing authentication, zero-config setup, and tools to list, query, create notebooks, and manage sources.63MIT
- AlicenseAqualityDmaintenanceAn MCP server that enables Claude to interact with Google NotebookLM via browser automation, allowing management of notebooks, adding sources, and querying information.7MIT
- AlicenseNot gradedqualityCmaintenanceMCP server for querying Google NotebookLM notebooks, enabling AI assistants to list notebooks, read sources, and ask questions about them.MIT