Base44 MCP server

Create and manage Base44 projects, edit an app’s code in its sandbox, and call the Base44 platform API from AI assistants such as ChatGPT, Claude, and Cursor

The Base44 MCP server exposes your Base44 account to any MCP-compatible AI assistant. Once connected, you can describe what you want to build or change and the AI will create or update projects on your behalf.

To connect, you must authenticate with OAuth. During that flow you choose which workspace to grant access to, and Base44 only acts within that workspace, with your explicit permission.

You need a Builder plan or higher to connect an AI assistant through the Base44 MCP server.

Connect your AI assistant

The server is available at:

https://app.base44.com/mcp

Add it to your AI tool’s MCP configuration:

"base44": {
  "type": "http",
  "url": "https://app.base44.com/mcp"
}

After adding the configuration, your AI tool will prompt you to sign in to Base44 through the standard OAuth flow. You only need to do this once.

Some tools require a restart to pick up new MCP configurations.

Choose a workspace

During the OAuth consent step, you choose which workspace the connection targets. Previously every connection defaulted to your personal workspace. Now you can point an assistant at a team’s shared workspace instead.

Every action the assistant takes, including creating, listing, and editing projects, runs in the workspace you pick, and the connection cannot reach any other workspace. The workspace is fixed for the life of the connection. To target a different workspace, connect again and select it during consent.

What your assistant can do in a workspace

Your assistant acts as you, so it inherits your role in the selected workspace. What it can do follows the same rules that apply to you in the Base44 editor.

• Creating projects: You need create permission in the workspace. In a shared workspace, a Viewer or Guest cannot create projects.
• Editing projects: You need write access to the project. A read-only Viewer cannot make changes, even to a project they can open.
• Reading projects: Listing projects, reading entity schemas, and querying records follow the same read access you already have.

SSO-enforced workspaces

Base44 re-checks your workspace membership on every request and every token refresh, so a connection only keeps working while your access does.

If the selected workspace enforces SSO, that check is stricter. Removing you from the workspace ends the connection right away, and rotating the workspace’s SSO configuration ends it until you connect again.

Available tools

The server exposes 3 kinds of tools. Project tools create and manage projects, and hand code changes to the app editor’s AI to make. Sandbox tools read, write, and run commands directly in the app’s sandbox, for when you want your assistant to write the code itself. Platform API tools reach the rest of the Base44 platform, such as deploys, analytics, and secrets, through its API.

Project tools

Tool	What it does
create_base44_app	Create a new project from a text description
edit_base44_app	Make changes to an existing project
list_user_apps	List your projects, optionally filtered by name
list_branches	List a project’s branches, with names, IDs, and status
list_entity_schemas	Get the entity schemas for a project
create_entity_schema	Define a new entity schema for a project
update_entity_schema	Replace the fields of an existing entity schema
query_entities	Query records from an entity in a project
create_entities	Create records for an entity in a project
update_entities	Update records for an entity in a project
list_connectors	List a project’s connectors and which ones are already connected
initiate_connector_connection	Connect a third-party connector for a project
import-claude-design-from-url	Import a design from a public URL and start building a project from it

update_entity_schema replaces the entity’s field definitions with the ones you send. Any field or constraint you don’t include is removed, not left as-is.

Sandbox tools

Tool	What it does
list_directory	List files and directories in the app’s sandbox
read_file	Read file contents from the app’s sandbox
grep	Search sandbox files for a pattern
write_file	Create or overwrite a file in the sandbox
edit_file	Apply exact string replacements to a file in the sandbox
run_command	Run a shell command in the sandbox
start_checkpoint	Start a checkpoint session before a change, which create_checkpoint finishes. Rolling out gradually, so it may not be available in your account yet
create_checkpoint	Save a restore point in the app’s version history, with an optional name

Sandbox tools that change the app need the sandbox:write permission, which you grant at the OAuth consent step. See Cloud sandbox for how the sandbox works, its permissions and limits, and the equivalent base44 sandbox CLI commands.

Changes your assistant makes with sandbox tools are saved to the app right away, but only checkpoints appear in the app’s version history. A Restore or Revert in the app editor rolls the app back to the last checkpoint and discards everything written after it. Ask your assistant to call create_checkpoint when it finishes a change.

Platform API tools

These tools reach Base44 platform capabilities that don’t have a dedicated MCP tool, such as deploys, checkpoints, published and preview URLs, analytics, secrets, custom domains, workflows, SEO, and Google Ads. Your assistant finds an operation in the catalog, reads its contract, and then calls it.

Tool	What it does
base44_readme	Get an overview of the available tools and how to use them together. Assistants call it once at the start of a conversation
list_api_catalog	List the Base44 platform API operations the assistant can call, grouped by area
describe_api	Get the parameters, request body, and response for up to 10 platform API operations
execute_api	Call a platform API operation by its method and path

For entity records, your assistant uses query_entities, create_entities, and update_entities rather than the platform API.

To find and call the same operations from your terminal, use the equivalent base44 api CLI commands.

Work on a branch

The edit_base44_app tool and the sandbox tools work on the app’s main branch unless you name another one. Your assistant gets each branch’s ID from list_branches and passes it to target that branch.

Internal tools

Specific MCP clients use 2 more tools internally to poll build progress. They aren’t meant to be called directly.

Tool	What it does
get_app_status	Check whether a project build is complete, or paused on a question you need to answer in the app editor
get_app_preview_url	Get the preview link for a project once its build is ready

Example prompts

Once connected, you can ask your AI assistant things like:

• “Create a Base44 project for tracking job applications with status, company name, and interview notes.”
• “Edit my CRM project to add a follow-up date field to the Contacts entity.”
• “List all my Base44 projects.”
• “Add 10 sample products to my inventory project so I can test the UI.”
• “Query the first 10 orders from my e-commerce project that have status pending.”
• “Publish my inventory project and send me its live URL.”

The AI interprets your request and calls the appropriate tool. For create_base44_app and edit_base44_app, the build runs in the background and you get a link to the Base44 editor when it’s ready.

See also

• Cloud sandbox: How the sandbox tools work, and what your assistant can change in an app
• Bring your own agent: Connect your own coding agent to an app’s sandbox
• Docs MCP server: Let AI assistants search Base44 documentation directly
• App MCP server: Let AI assistants connect to and use an app you built
• Skills: Reusable instructions that teach AI coding agents how to perform Base44-specific tasks
• Entities: Learn about the data model that list_entity_schemas and query_entities operate on