About notion-full-mcp
notion-full-mcp is a Notion server for MCP clients that supports page reading, editing, and recovery. It also handles tables and files, with annotations identifying read-only, additive, and destructive tools.
Key features
- 01Recursively read page content
- 02Edit blocks and table cells
- 03Snapshot and rollback restore
- 04Upload and download Notion files
- 05Move, copy, and restore pages
Use cases
- 01Inspect pages and nested content
- 02Change page text or table rows
- 03Back up and restore page content
- 04Insert files into Notion pages
MCP support
It exposes 33 tools for Notion pages, blocks, databases, comments, users, and files to MCP clients. Tool annotations identify read-only, additive, and destructive operations.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
uvx notion-full-mcpNOTION_API_TOKENSecretYour Notion integration API token (starts with ntn_)NOTION_MCP_FILE_ROOTOptional directory that uploads, downloads and snapshots are confined toLocal servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.
Tools
33Saved tool list · updated 2d ago
| Parameter | Type | Required | Description |
|---|---|---|---|
| query | string | Yes | |
| max_results | integer | No |
Details
From the project READMEnotion-full-mcp
Notion MCP server with deep page reading, surgical block editing, safe snapshot/restore, file uploads (Notion-hosted images and PDFs), table row manipulation, page move and copy, and page restore from trash.
What's new in v0.5.1
- Tool annotations. Every tool declares whether it is read-only, additive or destructive, so MCP clients can decide what to run without asking.
- MCP Registry. The server is published to the official MCP Registry as
io.github.cphoskins/notion-full-mcp.
What's new in v0.5.0
- Restore keeps the page in order. Subpages, inline databases and Notion-hosted files that are still on the page stay untouched, and the restored content is written around them in snapshot order.
- Restore brings items back from trash. A subpage, database or hosted file deleted after the snapshot is restored from Notion's trash with its ID, comments and history intact.
- Snapshots can carry files.
snapshot_pagewithinclude_filesdownloads Notion-hosted images and files beside the JSON, and restore uploads them again in place. - Fix: inserting more than 100 blocks after a block now keeps them in order.
What's new in v0.4.0
- Restore can no longer lose content.
restore_page_from_snapshotwrites the snapshot first and removes the old blocks only after that succeeds. A failed restore rolls back and leaves the page as it was. Subpages, inline databases and Notion-hosted files are never deleted. move_pageworks. It uses Notion's move endpoint, so the page keeps its ID, content, comments and subpages.copy_page_to_parentreports what it could not copy. Uncopyable block types are listed inskipped, and the original is archived only when everything was copied.- Large and deep content. Lists nested more than two levels, tables over 100 rows and text over 2,000 characters are split across requests automatically.
- Mentions and equations survive edits. Find/replace, restore and copy keep user, page, database and date mentions and inline equations.
- Better errors and rate-limit handling. Errors carry Notion's own code and message. Rate-limited requests are retried using
Retry-After. - Database rows have titles. Search, page info and create return the title whatever the title column is called.
- More block types. Equation, bookmark, embed and externally hosted media blocks round-trip through snapshot/restore and copy.
- Optional file boundary. Set
NOTION_MCP_FILE_ROOTto confine uploads, downloads and snapshots to one directory.
What's new in v0.3.0
- Page restore from trash.
restore_pageexplicitly unarchives a page by sendingarchived=False. This is the counterpart todelete_blockon a page and works around the intentional safety guard inupdate_pagethat blocksarchived=False. Cascaded child archives are restored automatically when the parent is restored. - Destructive page copy across parents.
copy_page_to_parentis the workaround for Notion's public API not supporting parent changes. Recreates the page at the target parent with the same title, icon, and content, then archives the original. Handles paragraphs, headings, lists, tables, dividers, code blocks, and image/file blocks. Refuses to run if the source has nestedchild_pagedescendants (copy those manually). move_pagestopped pretending. Notion'sPATCH /pages/{id}endpoint silently ignores parent changes, so the tool returned an error instead. (v0.4.0 implements a real move.)
What's new in v0.2.0
- Notion file uploads. Upload images and files directly to Notion's own workspace storage instead of relying on an external image host. Works for PNG, JPEG, SVG, GIF, PDF, and any other Notion-supported attachment type.
- Image block helpers.
upload_image_as_blockuploads a file and inserts it at a position in one call.replace_image_blockswaps an existing image's content while preserving its position in the parent.download_image_blockpulls any image block's file to disk. - Table row manipulation.
insert_table_rowadds a row to an existing table with automatic width validation.update_table_cellreplaces the text of a single cell. No more rebuild-the-whole-table for small edits.
Install
pip install notion-full-mcp
# or
uv pip install notion-full-mcp
Quick Start
export NOTION_API_TOKEN="your-token-here"
notion-mcp
Claude Code Integration
Add to ~/.claude.json under mcpServers:
{
"notion-full": {
"type": "stdio",
"command": "uvx",
"args": ["notion-full-mcp"],
"env": {
"NOTION_API_TOKEN": "your-token-here"
}
}
}
Or for local development:
{
"notion-full": {
"type": "stdio",
"command": "uv",
"args": ["run", "--directory", "/path/to/notion-full-mcp", "notion-mcp"],
"env": {
"NOTION_API_TOKEN": "your-token-here"
}
}
}
Tools (33)
Search & Pages
| Tool | Description |
|---|---|
search_pages | Search pages by title |
search | Search pages and/or databases |
get_page_info | Get page metadata (title, parent, dates, URL) |
create_page | Create a new page under a parent page or database |
update_page | Update a page's title, icon, or archive status |
move_page | Move a page under a new parent page, keeping its ID and content |
restore_page | Restore a page (and its cascade-archived children) from trash |
copy_page_to_parent | Recreate a page under a new parent page or database. Archives the original only when every block was copied |
Reading
| Tool | Description |
|---|---|
read_page_deep | Read page with all nested children (recursive) |
list_block_children | List direct children of a block or page |
get_block | Get a single block by ID |
Editing
| Tool | Description |
|---|---|
update_block_text | Replace all text in a block |
find_replace_in_block | Find/replace text preserving formatting |
update_block_rich_text | Update with full rich text specification |
Inserting
| Tool | Description |
|---|---|
insert_text_after | Insert a text block after a specific block (paragraph, heading_1/2/3, bulleted_list_item, numbered_list_item, quote, callout, toggle, to_do, code, divider) |
insert_blocks_after | Insert multiple blocks after a specific block |
append_blocks_to_page | Append blocks to the end of a page |
Deleting
| Tool | Description |
|---|---|
delete_block | Delete a block by ID |
Comments
| Tool | Description |
|---|---|
create_comment | Add a comment to a page |
list_comments | List all comments on a page or block |
Databases
| Tool | Description |
|---|---|
get_database | Get database metadata and schema |
query_database | Query a database with filters and sorts |
Users
| Tool | Description |
|---|---|
get_self | Get the current authenticated user/bot |
list_users | List all workspace users |
Snapshot / Restore
| Tool | Description |
|---|---|
snapshot_page | Save complete page structure as JSON. include_files also saves Notion-hosted images and files |
restore_page_from_snapshot | Restore page content from a snapshot. Writes first, rolls back on failure, keeps subpages and databases in place, restores deleted ones from trash and re-uploads saved files |
Media & Files
| Tool | Description |
|---|---|
upload_file | Upload a file to Notion's own workspace storage (returns file_upload_id) |
upload_image_as_block | Upload an image and insert it as a new image block in one step (after a block or appended to a page) |
replace_image_block | Replace an existing image block with a newly uploaded file, preserving its position |
download_image_block | Download the image file from an image block to disk — works for external URLs, Notion-hosted files, and file_upload references |
upload_file_as_block | Upload any file (PDF, attachment) and insert it as a file block |
Tables
| Tool | Description |
|---|---|
insert_table_row | Insert a new row into an existing table with automatic width validation |
update_table_cell | Replace the text of a single cell in a table row |
Tool annotations
Every tool carries MCP annotations so a client can decide what is safe to run without asking:
- Read-only (11): search, page and block reads, comments list, database reads and user lookups.
- Adds without loss (13): inserts, appends, uploads, page and comment creation,
move_page,restore_page, and the two tools that write local files (snapshot_page,download_image_block). - Marked destructive (9):
update_block_text,update_block_rich_text,find_replace_in_block,update_table_cell,replace_image_block,delete_block,update_page,copy_page_to_parentandrestore_page_from_snapshot. They overwrite, delete or archive existing content.
Block support and limits
- Round-trip (snapshot/restore and copy): paragraph, headings (including toggleable), bulleted and numbered lists, to-do, quote, callout, toggle, code, divider, table, columns, table of contents, equation, bookmark, embed, and externally hosted image, file, video, PDF and audio blocks.
- Never recreated from scratch: subpages, inline databases and synced blocks. Copy reports them as
skipped. - Restore of those blocks, and of Notion-hosted files:
- Still on the page: left untouched, with restored content written around them in snapshot order (
preserved). One that was dragged out of snapshot order stays where it is (out_of_order). - File saved with the snapshot: uploaded again and recreated in place (
uploaded). Files over 20 MB, or over the workspace upload limit, are not uploaded. - In Notion's trash: brought back as the same block (
restored_from_trash). Notion returns it to the end of the page, so a block that is still on the page and came later in the snapshot ends up above it and is reported asout_of_order. - Gone from trash, with no saved file: cannot be restored (
skipped). - Why
out_of_orderblocks are not moved (by design): the Notion API has no way to reposition a block. The only way to move a subpage or database is to trash it and restore it, which sends it to the end of the page and, if it failed midway, would leave it in trash. Restore never deletes something it cannot recreate, so it reports the block and leaves it for you to drag into place.
- Still on the page: left untouched, with restored content written around them in snapshot order (
- Snapshots: record subpages and inline databases as blocks, without their own content. Notion-hosted files are saved only with
include_files; otherwise the snapshot holds links that expire within the hour. - Queries:
query_databasereturns at most 500 rows and setstruncatedwhen more matched. - Find/replace: matches within one formatting run. A phrase that spans a bold or link boundary is reported as not found.
- API version: requests use Notion API version
2022-06-28.
Configuration
| Variable | Required | Purpose |
|---|---|---|
NOTION_API_TOKEN | Yes | Integration secret |
NOTION_MCP_FILE_ROOT | No | When set, uploads, downloads and snapshot files must be inside this directory |
Development
uv run pytest --cov=notion_mcp # unit tests, no credentials needed
A live check runs every tool against a real workspace. It works inside one scratch page under the parent you name and moves it to trash afterwards:
export NOTION_API_TOKEN="your-token-here"
export NOTION_TEST_PARENT_PAGE="page-id-shared-with-the-integration"
uv run python scripts/live_check.py
Notion API Token
- Go to https://www.notion.so/my-integrations
- Create a new integration
- Copy the token (starts with
ntn_) - Share the pages/databases you want to access with the integration
License
MIT
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.