# MCP Tools Reference


147 tools, grouped by what they touch. You never call these by hand — your assistant picks them. This page is here so you know what to ask for, and what a key profile is letting through.

:::note
**Naming conventions.** Tools starting with `list_`, `get_`, `search_`, `ask_`, `compare_` or `is_` only read. Tools that change something are labelled **WRITE**, and irreversible ones are labelled **DESTRUCTIVE** — good clients ask you to confirm those.
:::

## Connection & account

| Tool | What it does |
| --- | --- |
| `get_status` | Whether the extension is connected, which Google account calls run as, and the extension version |
| `list_accounts` | Every Gemini Notebook account available to this Kortex user |
| `get_usage` | Your **Kortex** quota usage for imports, exports, prompts and source views |
| `get_notebooklm_usage` | **Google's** own account state — the Google-side limits that account has hit. A different ceiling from the one above, and the one that stops an Audio Overview generating |
| `get_account_metadata` | The raw Kortex metadata blob for the active account |
| `get_job` | The result of a long-running tool. Anything that may outlast the request hands back a job id instead of timing out; your assistant polls this until it is done |

## Notebooks & sources

| Tool | What it does |
| --- | --- |
| `list_notebooks`, `get_notebook` | List notebooks, or read one with its full source list |
| `list_sources`, `list_all_sources` | Sources in one notebook, or across your whole library. `notebook_id` is optional — omit it and you get every source, each row carrying the notebook it belongs to, instead of listing notebooks and asking once per notebook |
| `get_source_content` | Read a source page by page, as text or HTML, with a cursor for the next page |
| `get_source_summaries` | Gemini Notebook's own summary and key topics for your sources, without reading a word of the source text. Omit the ids and it does the whole notebook in one call — the cheap way to find out what is in a library |
| `get_notebook_guide` | The notebook's own summary plus the questions Gemini Notebook suggests about it, each with the ready-made prompt behind it |
| `get_formatted_content` | Return a source or all notes as Markdown or plain text, with no download |
| `create_notebook`, `rename_notebook`, `duplicate_notebook` | Create, rename or clone a notebook |
| `merge_notebooks`, `move_sources_between_notebooks`, `compare_notebooks` | Combine notebooks, move sources between them, or diff two by source title |
| `rename_source`, `is_source_ready`, `refresh_notebook_data` | Rename a source, check ingestion status, force a data refresh |
| `delete_sources`, `delete_notebook`, `delete_bad_sources` | **Destructive.** Permanent deletion of sources, a notebook, or every failed/empty source |
| `set_notebook_language`, `list_supported_languages` | Set the language a notebook answers and generates in |
| `share_notebook`, `open_notebook` | Read sharing, publish to anyone with the link, or invite someone by email as viewer or editor; open a notebook in your browser |

`compare_notebooks` is one of the tools that answers with a panel instead of a list, because the useful part of a diff is deciding what to do about it:

<img src="/docs/img/docs/mcp-panel-compare.png" alt="Comparing two notebooks" width="560" />

Tick the sources that are missing from one side, copy or move them across, or merge both into a new notebook. Every button there runs the real tool.

## Asking questions

| Tool | What it does |
| --- | --- |
| `ask_question` | A grounded question against one notebook. Returns the answer, citations and a `session_id` for follow-ups |
| `ask_library` | The same question across up to three notebooks, chosen by id, tag or collection |
| `list_chat_sessions`, `reset_chat_session`, `close_chat_session` | Manage the durable conversation threads behind `session_id` |
| `batch_to_vault` | Run a long list of questions against a notebook and keep every answer with its citations |
| `save_chat_to_note` | Save the latest answer, citations included, as a native Notebook note |

## Adding sources

| Tool | What it does |
| --- | --- |
| `add_url_source`, `add_text_source` | Add web pages, YouTube links, or pasted text |
| `add_google_doc`, `sync_google_docs` | Link a Google Doc as a live source and re-read it later |
| `save_url`, `save_snippet`, `fetch_url` | Fetch and clean a public page, save a passage, or read a page without saving it |
| `start_file_upload`, `upload_file_chunk`, `finish_file_upload`, `cancel_file_upload` | Upload a local PDF, DOCX or TXT in chunks (50 MiB max, one-hour staging) |
| `import_youtube` | Import up to 50 videos from a channel, playlist or search page |
| `import_pipeline` | Bulk-import from a link list, CSV, RSS/Atom feed, or a bounded site crawl |
| `import_social_post`, `export_chat_to_notebook` | Import a Reddit post, or an AI chat conversation from a supported platform |
| `research_sources` | Start, poll and import source discovery — Fast, or Deep Research for a thorough pass that takes minutes |

## Studio artifacts & podcasts

| Tool | What it does |
| --- | --- |
| `generate_audio_overview`, `generate_artifact` | Start an Audio Overview, study guide, video overview or data table |
| `list_artifacts`, `export_artifact`, `delete_artifact` | Check generation status, export to media/Markdown/CSV/Docs/Sheets, or delete. `list_artifacts` takes an optional `notebook_id`: omit it to sweep every notebook in one call, with anything unreadable reported rather than failing the sweep |
| `get_artifact_content` | The cards, questions or nodes inside a flashcard deck, quiz or mind map — the parts the list view does not carry |
| `open_artifact_studio` | Open the interactive Studio view in your client, to pick sources and options before generating |
| `rename_artifact` | Retitle a Studio artifact, so a column of identical "Briefing Doc" rows becomes navigable |
| `revise_slide_deck` | Revise named slides of an existing deck. Expected to produce a new deck and leave the original alone; the reply tells you which happened |
| `list_podcasts`, `create_podcast`, `update_podcast`, `delete_podcast` | Manage Kortex podcasts built from Audio Overviews |
| `add_podcast_episode`, `remove_podcast_episode`, `reorder_podcast_episodes` | Manage the episode list |
| `get_podcast_feed_url`, `download_podcast_zip` | Get the public RSS URL, or download every episode as a ZIP |

A finished flashcard deck or quiz comes back as something you work through, not as a wall of JSON:

<img src="/docs/img/docs/mcp-panel-flashcards.png" alt="Flashcard panel in the chat" width="520" />

<img src="/docs/img/docs/mcp-panel-quiz.png" alt="Quiz panel in the chat" width="520" />

The podcast tools do the same — the feed's details, its episodes and their order in one view, with the RSS URL a click away:

<img src="/docs/img/docs/mcp-panel-podcast.png" alt="Managing a podcast feed" width="520" />

## Notes

| Tool | What it does |
| --- | --- |
| `list_notes` | Native Notebook notes and generated artifacts |
| `create_notebooklm_note`, `update_notebooklm_note`, `delete_notebooklm_note` | Write to Gemini Notebook's own Studio notes |
| `convert_note_to_source` | Turn a note into a new source in the same notebook |
| `list_kortex_notes`, `save_kortex_note`, `delete_kortex_note` | Kortex's own notes, attached to a notebook in the dashboard |
| `list_note_templates`, `save_note_template`, `delete_note_template` | Reusable note bodies |

## Organising your library

| Tool | What it does |
| --- | --- |
| `list_collections`, `create_collection`, `rename_collection`, `move_collection`, `pin_collection`, `delete_collection` | Kortex collections (the folders in your dashboard) |
| `move_notebook_to_collection` | File a notebook into a collection |
| `list_notebooklm_collections`, `create_notebooklm_collection`, `migrate_collections_to_notebooklm` | Gemini Notebook's own native collections, and a one-way migration from Kortex folders |
| `update_notebooklm_collection`, `delete_notebooklm_collection` | Rename a native collection, set its emoji, add or remove notebooks, or delete the grouping. The notebooks themselves are never deleted |
| `list_tags`, `tag_notebook`, `untag_notebook`, `delete_tag`, `set_favorite` | Notebook tags and favourites |
| `list_source_organizers`, `create_source_folder`, `assign_source_to_folder`, `tag_source`, `bulk_assign_source_organizers` | Source tags and folders, one at a time or in bulk |
| `delete_source_folder`, `delete_source_tag` | **Destructive.** Remove a source folder or tag everywhere |
| `list_source_labels`, `create_source_label`, `update_source_label`, `delete_source_labels` | Gemini Notebook's native source labels, with emoji |
| `auto_label_sources` | Let Gemini Notebook invent labels for a notebook and file every source under them. Its own auto-organise, not a Kortex guess |
| `set_source_highlight`, `set_source_note` | Private colour highlights and notes on a source |
| `list_source_views`, `save_source_view`, `delete_source_view` | Named subsets of sources you can re-select later |
| `list_notebook_templates`, `save_notebook_as_template`, `apply_notebook_template` | Reusable presets of tags and collections |
| `list_source_templates`, `save_source_as_template`, `apply_source_template` | Reusable presets of source tags and a folder |

## Doing many things at once

Every Kortex call is a job the extension runs, and jobs run one at a time — Gemini Notebooks tolerates concurrent requests poorly. So twenty separate calls are twenty round trips, and your assistant cannot speed that up by issuing them in parallel: the server accepts at most three at once and the extension still works through them in order.

`bulk` is the way round it. One call, many targets, run concurrently _inside_ that single job at the safe rate for the work — five reads, three writes, or two generations at a time.

<img src="/docs/img/docs/concept-bulk.svg" alt="One at a time, or all at once" width="760" />

| Tool | What it does |
| --- | --- |
| `bulk` | Run one tool over many targets in a single call. Reach for it whenever the same tool is about to be called more than about three times |

**Creating and filling in one step.** A `create_notebook` entry may carry its own `urls`, so each item creates the notebook _and_ adds its sources. The new notebook's id never has to come back to your assistant and go out again:

```
bulk(tool: "create_notebook", calls: [
  { title: "Ancient Rome",       urls: [...] },
  { title: "Deep Sea Creatures", urls: [...] }
])
```

**If it does not all fit.** A long batch stops cleanly at the request budget rather than timing out. Unfinished entries come back marked _"Not attempted"_ with a `task_id`; calling `bulk` again with that id carries on where it stopped.

**What it will not do.** `bulk` runs a fixed list of safe tools — creating notebooks, adding sources, tagging, renaming, generating artifacts and reading. It refuses every destructive tool by name, because a single confirmation covering twenty deletions is exactly what the confirm step exists to prevent. Delete one at a time, on purpose.

## Search, health & prompts

| Tool | What it does |
| --- | --- |
| `search_everything` | One search across notebooks, sources, notes, artifacts, collections, tags and podcasts |
| `get_library_health` | Bad sources, never-viewed sources, failed artifacts, notebooks near their cap |
| `dedupe_library` | A resumable duplicate scan across every notebook's sources |
| `semantic_library_metadata` | Build and maintain a searchable description of each notebook, so your assistant can pick the right one |
| `list_prompts`, `save_prompt`, `delete_prompt` | Your Kortex prompt library |
| `pick_prompt` | Choose a saved prompt and fill its placeholders, without pasting the text yourself |
| `list_prompt_folders`, `create_prompt_folder`, `delete_prompt_folder` | Prompt folders |

## Workflows

A workflow is a sequence of tool calls saved under a name, so a multi-step job you repeat becomes one request. Unlike an automation it runs when you ask, not on a schedule, and a step can wait for a generation to finish before the next one starts — which is what lets a later step use an Audio Overview an earlier step kicked off.

You do not write them in a config file. You do the job once by asking, then ask for it to be saved:

```
Save that as a workflow called "Competitor brief"
```

### What people save

| Workflow | The steps behind it |
| --- | --- |
| **"Run my competitor brief"** | Add this week's URLs to the notebook, wait until they finish indexing, generate a briefing doc, export it to Google Docs. |
| **"Set up a new course notebook for &#123;subject&#125;"** | Create the notebook, apply your notebook template, tag it, file it into the right collection. The `&#123;subject&#125;` is a declared input, so one workflow serves every course. |
| **"Run my weekly digest"** | Ask the same question across three notebooks, stitch the answers together, save the result as a note. |
| **"Turn this notebook into an episode"** | Generate an Audio Overview, wait for it to finish, add it to a podcast as an episode, return the feed URL. |
| **"Ingest today's reading"** | Import a list of URLs, delete anything that failed to extract, then report what actually landed. |

**Inputs.** A workflow can declare inputs — a notebook, a subject, a date — so the same saved sequence works on different targets. Ask for one without its required inputs and it tells you what is missing rather than running a half-configured job.

**Workflows can be scheduled.** A saved workflow becomes an automation action, so anything you can do by hand you can put on a clock. See [Workflows & Schedules](/docs/mcp-workflows).

### Tools

| Tool | What it does |
| --- | --- |
| `list_workflows` | Your saved workflows, with their steps and declared inputs |
| `create_workflow`, `update_workflow` | **WRITE.** Save or change a named sequence of steps |
| `run_workflow` | **WRITE.** Run one, optionally with inputs. Returns a run id |
| `get_workflow_run` | Progress of a run: each step's tool, whether it succeeded, and the final result |
| `delete_workflow` | **DESTRUCTIVE.** Remove a saved workflow |

## Automations

Eight tools create and manage standing rules that Kortex runs on its own — `list_automations`, `create_automation`, `update_automation`, `set_automation_enabled`, `delete_automation`, `run_automation`, `list_automation_runs` and `run_scheduled_automations`. Because a rule keeps acting after the conversation ends, they have [their own page](/docs/mcp-workflows).