Connect AI assistants to your Baserow databases using the Model Context Protocol (MCP). The Baserow MCP server lets you use ChatGPT, Claude, Cursor, Windsurf, and other MCP-compatible AI tools to search, read, create, update, and delete database records using natural language.
Whether you want to query a database with ChatGPT, manage records with Claude, or work with your data through Cursor, Baserow MCP provides a direct connection between your database and AI tools.
You don’t need to write SQL queries or build custom API integrations to get started.

The Model Context Protocol (MCP) is an open standard that allows AI applications to connect to external tools, databases, and services.
Baserow’s native MCP server connects AI assistants directly to your Baserow workspaces. Once connected, an AI assistant can discover your databases, inspect table schemas, search records, and create, update, or delete rows based on your instructions.
Instead of manually navigating tables or writing API requests, you can describe what you want to do in natural language.
For example, you can ask an AI assistant to:
The Baserow MCP server acts as a connection between your Baserow workspace and an AI assistant.
When you send a request, the assistant uses the available MCP tools to discover your database structure, retrieve relevant records, and perform the requested operations.
There are two ways to connect Baserow to an AI assistant:
Direct integrations
Connect Baserow through an AI assistant’s app or connector directory. Sign in with your Baserow account, select a workspace, and choose which tools the assistant can use.
Baserow currently offers direct integrations with:
Custom MCP connections
Connect Baserow to other MCP-compatible applications, such as Cursor and Windsurf, using an MCP server URL and the configuration required by your AI client.
Baserow MCP supports the following database operations:
| Operation | Available actions |
|---|---|
| Discover databases | List databases and tables |
| Inspect database structure | Retrieve table schemas and field definitions |
| Read data | Search, retrieve, and paginate through table rows |
| Create data | Create new rows |
| Update data | Modify existing rows |
| Delete data | Delete rows from tables |
The current row-management integration can read database, table, and field information, but it only changes rows. You cannot create or modify databases, tables, or fields through these tools.
Deleted rows go to Baserow’s trash, where they can be restored for three days by default.
Baserow MCP uses authentication and workspace-level access to control how AI assistants interact with your data.
When connecting through the direct ChatGPT or Claude integrations:
The assistant can only use the tools and workspace authorized during sign-in.
For custom MCP connections, access is managed through the endpoint configuration and authentication method. Treat your MCP URL as a credential and store it securely.
Only grant permissions required for your intended use case. Review the permissions before authorizing a connection, particularly when enabling operations that modify or delete records.
Connect your Baserow workspace to ChatGPT using the Baserow app. Once connected, ChatGPT can access your authorized workspace and help you search, create, update, and delete records using natural language.
You don’t need to configure an MCP endpoint manually or write JSON configuration.
Open the Baserow app in ChatGPT.
Select the option to connect the app.
Review the connection details and select Continue to Baserow.
ChatGPT will redirect you to Baserow to authorize the connection.
Sign in using your Baserow credentials.
If you are already signed in, you may be redirected directly to the authorization screen.
Choose the Baserow workspace you want to connect to.
Each connection is associated with one workspace. If you need to access another workspace, you will need to establish a separate connection.
Choose which tools ChatGPT can use in your workspace.
The authorization screen groups permissions into categories, including:
Databases
Automations
The authorization screen may also display permissions for reading and managing workflows and workflow nodes.
Only enable the permissions you need. The available tools depend on the capabilities exposed by the connected Baserow integration.
Only read tools are selected by default. You can grant additional permissions depending on what you want ChatGPT to do.
Review your selections and click Allow.
After authorization, you’ll be redirected back to ChatGPT.
Your Baserow connection is now available.
You can ask ChatGPT to find information, search records, create new rows, or update existing data, depending on the tools you authorized.
Example prompts:
Tip: Be specific about the database and table you want to work with. For changes affecting multiple records, ask ChatGPT to show you which rows will be changed before proceeding.
Connect your Baserow workspace to Claude using the Baserow connector in Claude’s Connectors Directory.
Once connected, Claude can discover your databases, inspect field definitions, search and retrieve records, and create, update, or delete rows through natural language instructions.
You don’t need to configure a local MCP server or manually add JSON configuration for this direct integration.
Open Claude and navigate to Customize → Connectors.
Select the option to browse available connectors.
Find Baserow in the directory.
Alternatively, visit the Claude Connectors Directory and search for Baserow.
Select the Baserow connector and click Connect.
Claude will display the connector information, including its developer and connection URL.
Baserow is listed as a community connector. Claude displays a notice explaining that community connectors undergo automated reviews but aren’t verified by Anthropic.
Before connecting, check that the connector is published by Baserow and that its connection URL is:
https://api.baserow.io/mcp
Select Connect to continue.
Claude may display a notification about updates to its Consumer Terms and Privacy Policy.
Review the information and accept the updated terms to continue.
Claude will redirect you to Baserow.
Sign in with your Baserow account to authorize the connection.
Choose the Baserow workspace you want Claude to access.
The connection will be limited to the selected workspace.
Review the available Baserow tools and select which operations Claude can perform.
The authorization screen organizes permissions into categories.
Databases
Automations
The authorization screen may also display permissions for reading and managing workflows and workflow nodes.
The Baserow connector currently exposes the following row-management tools:
list_databaseslist_tablesget_table_schemalist_table_rowscreate_rowsupdate_rowsdelete_rowsOnly read tools are enabled by default. Enable additional permissions only if you want Claude to perform those actions.
Click Allow to authorize the connection.
Once authorization is complete, Claude will show Baserow as a connected service.
You can now use Claude to interact with your Baserow workspace.
Example prompts:
Tip: Ask Claude to inspect your table schema before performing operations. This helps it use the correct field names and understand your database structure.
You can also connect Baserow to AI applications that support custom MCP servers.
This method is useful when your AI client doesn’t offer a direct Baserow integration in its app directory.
Baserow will generate a unique MCP URL for your connection.
Copy the URL and store it securely.
After creating your endpoint, click More Details to view the available configuration options.
Select your preferred AI client and follow its setup instructions.
{
"mcpServers": {
"Baserow MCP": {
"url": "YOUR_MCP_URL_HERE"
}
}
}
Replace YOUR_MCP_URL_HERE with the MCP URL copied from your Baserow settings.
mcp_config.json.{
"mcpServers": {
"Baserow MCP": {
"serverUrl": "YOUR_MCP_URL_HERE"
}
}
}
Replace YOUR_MCP_URL_HERE with the MCP URL copied from your Baserow settings.
claude_desktop_config.json.{
"mcpServers": {
"Baserow MCP": {
"command": "npx",
"args": [
"mcp-remote",
"YOUR_MCP_URL_HERE"
]
}
}
}
Replace YOUR_MCP_URL_HERE with the MCP URL copied from your Baserow settings.
Once configured, your AI client can connect to Baserow and use the tools exposed by your MCP endpoint.
Follow your client’s instructions to complete authentication and verify the connection.
Note: Custom MCP configurations and authentication requirements may differ between AI clients. For self-hosted Baserow installations, additional MCP authentication configuration may be required.
Baserow exposes tools that allow AI assistants to discover database structures and work with records.
| Tool | Description |
|---|---|
list_databases |
Lists databases available in the connected workspace. |
list_tables |
Lists tables in a database. |
get_table_schema |
Retrieves a table’s fields and their definitions. |
list_table_rows |
Searches and retrieves rows from a table. |
create_rows |
Creates new rows in a table. |
update_rows |
Updates existing rows. |
delete_rows |
Deletes rows from a table. |
For reliable results, an AI assistant should follow this sequence:
list_databases to find the available databases.list_tables to identify the relevant table.get_table_schema to understand the table’s fields.list_table_rows to retrieve relevant records.Use the database and table IDs returned by the discovery tools. Never guess IDs or field names. Use field names exactly as returned by the table schema.
Use list_table_rows with a search term instead of retrieving an entire table whenever possible.
The tool returns:
count: The total number of matching rows.size rows.If the total number of matching records is larger than the number returned, retrieve subsequent pages or narrow the search before concluding that a record doesn’t exist.
Search may not cover link-to-table, boolean, collaborator, and duration fields. If searching for these values returns no results, retrieve the relevant rows and compare the field values directly.
Before using update_rows or delete_rows, the assistant should show you which rows will be affected.
Large changes should be processed in batches of no more than 200 rows.
Deleted rows go to Baserow’s trash, where they can be restored for three days by default. There is no undo tool available through the MCP connection.
The current row-management tools can read database structures and manage rows, but they cannot:
Use the Baserow interface to make structural changes to your databases.
Your Baserow MCP connection gives an AI assistant access to the tools and workspace you authorized. The level of access depends on the connection method and permissions you grant.
For custom MCP endpoints, your MCP URL contains credentials that grant access to the endpoint.
Do:
Don’t:
For direct ChatGPT and Claude integrations, manage authorization through the connected app and your Baserow MCP settings.
If list_databases returns no results, the connection may point to a workspace that doesn’t contain any databases.
Check that you selected the correct workspace when authorizing the connection.
Check the permissions you granted when connecting Baserow.
If the required row tool wasn’t authorized during sign-in, you need to authenticate again and enable the relevant permission.
For direct ChatGPT and Claude connections, a plain reconnect may retain the previous authorization.
To change the workspace or permissions:
The list_table_rows tool returns paginated results, so a search may not include every matching record in the first response.
Ask the assistant to check the total number of matching records and retrieve additional pages if necessary.
Some field types, including link-to-table, boolean, collaborator, and duration fields, may not be covered by search. In these cases, the assistant should retrieve the rows and compare the field values directly.
Each connection is associated with one workspace.
To work with multiple workspaces, establish separate connections and authorize each workspace.
No. The current row-management MCP tools can read database structures and manage rows, but they cannot create or modify databases, tables, or fields.
Use the Baserow interface to make structural changes to your databases.
Yes. You can disconnect direct integrations and manage your MCP connections under My Settings → MCP Server in Baserow.
For custom MCP endpoints, you can delete the endpoint in your Baserow MCP settings. If an endpoint URL is compromised, delete it and create a new one. The old URL will no longer work.
You can also disconnect integrations from your AI assistant’s connected apps or connectors settings.
MCP (Model Context Protocol) is an open standard that allows AI applications to connect to external tools, databases, and services.
An MCP server exposes tools that an AI assistant can use to retrieve information and perform actions in an external application.
The Baserow MCP server connects AI assistants to Baserow databases. It allows assistants to discover databases, inspect schemas, search records, and create, update, or delete rows using natural language instructions.
Baserow offers direct integrations with ChatGPT and Claude. You can also connect Baserow to other MCP-compatible applications, including Cursor and Windsurf, through custom MCP configurations.
No. The direct integrations use your Baserow account to authenticate and authorize access to a selected workspace. You don’t need to manually create an API token or copy an MCP endpoint URL.
Yes. Self-hosted Baserow can be connected using a custom MCP server configuration. Your installation must have MCP authentication enabled and be configured to accept connections from the AI client.
See the MCP OAuth setup documentation for configuration details.
Yes. You can ask an AI assistant to search records, create new rows, update existing information, or delete records using natural language.
The assistant uses your database schema and available MCP tools to translate your instructions into database operations.
Baserow MCP uses dedicated tools to interact with your databases rather than requiring you to write SQL queries.
You can describe the information you need in natural language, and the AI assistant can search and retrieve relevant records.
No. Baserow MCP supports both read and write operations.
Read-only tools are enabled by default when connecting. You can authorize additional tools to create, update, and delete rows.
You can use AI assistants to perform data-management tasks through MCP, including creating and updating records.
MCP connects an AI assistant to Baserow; it is not itself an automation workflow engine. For more advanced automated processes, explore Baserow Automations.
Yes. You can create multiple custom MCP endpoints within the same workspace. This is useful for organizing access by project, team member, or use case. Each endpoint has its own URL and access configuration.
If your custom MCP URL is compromised, immediately delete the endpoint in your Baserow MCP settings and create a new one. The old URL will no longer function.
No. You can describe the database operations you want to perform in plain English. The AI assistant uses the available MCP tools to translate your instructions into database operations.
AI interpretations depend on model quality and prompt clarity. Always verify critical operations, especially bulk updates or deletions.
The current row-management MCP tools support reading database structures and managing rows, but do not support creating or modifying databases, tables, or fields.
Still need help? Visit the Baserow Community or contact our support team.