WeWeb MCP

CANNOT USE PLUGINS

Please note that the WeWeb MCP does not expose ways for AI to build using plugin actions. Thus, if you are using a plugin such as Xano or Supabase, your AI will not be able to build using these respective plugin actions.

In the cases of Xano and Supabase, we recommend you switch to using the new Integrations for these providers, as the MCP does expose ways for your AI to build using integration actions.

If you still need to work with plugin actions from an MCP-connected AI, there is a makeshift workflow JSON workaround further down this page. It is not ideal, and we recommend Integrations whenever they cover your use case.

Use the WeWeb MCP server to let an external AI agent work on your WeWeb projects. Instead of building inside the editor with WeWeb AI, you can stay in the AI client you already use, such as ChatGPT, Claude, Cursor, Codex, or Antigravity, and ask it to create pages, manage your backend, edit workflows, and inspect your project through natural language.

The server is available at:

https://ai-api.weweb.io/v1/mcp

MCP trial on the Free plan

The WeWeb MCP server is available on a 14-day trial when you are on the Free seat plan. This gives you time to connect an AI client and try building with MCP without upgrading to a paid plan.

When the trial ends, you'll need to upgrade to a paid Pro or Partner seat plan if you want to keep using the MCP server. Without a Pro or Partner plan, your AI client can no longer call WeWeb tools through MCP.

Learn more about WeWeb pricing →

WeWeb recommends Claude Opus

For the best results today, we recommend using Claude Opus as the underlying model. In our testing, it follows WeWeb tooling more reliably and produces higher-quality output than other models. As the MCP server matures, we'll broaden the list of recommended models.

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI agents discover and call external tools in a consistent way. By exposing WeWeb as an MCP server, any compatible client can list the available WeWeb tools and invoke them to act on your project.

This is the inverse of a chat connector: rather than WeWeb calling out to an external service, an external agent calls into WeWeb to do the work for you.

How it works

1. Connect your AI client to the WeWeb MCP server using the steps below.
2. Authenticate. On first use, the client opens a browser window so you can sign in to WeWeb and authorize access.
3. Select a workspace and project. Ask the agent to list your workspaces and projects, then switch to the ones you want to work on.
4. Build and iterate. Describe what you want and let the agent call the relevant WeWeb tools.

CURRENTLY IN BETA

WeWeb MCP is currently in beta. Tool calls run against your real project and can create, edit, or delete resources. We recommend testing on a non-critical project first and reviewing changes as the agent works.

Supported clients

The following clients are supported today:

Client	Connection method	Guide
ChatGPT	Add a custom plugin with the server URL	Setup guide
Claude Desktop	Add a custom connector with the server URL, or edit the config file manually	Setup guide
Claude Code	Single terminal command	Setup guide
Cursor	MCP config entry	Setup guide
VS Code	MCP config entry, or add through MCP settings	Setup guide
Codex	Add a custom plugin with the server URL	Setup guide
Antigravity	Add an entry to mcp_config.json	Setup guide

Authenticate and manage tools

No matter which client you use, connecting to the WeWeb MCP server prompts you to sign in to WeWeb and authorize access. The first time you connect, or when your session expires, your client opens a browser window with a screen like this:

Review what access is being requested, confirm you are signed in to the correct WeWeb account, then click Authorize. Until you approve, the agent cannot call WeWeb tools on your behalf.

What are MCP tools?

An MCP server exposes tools: discrete actions the agent can run, such as creating a page, fetching table view data, or editing a workflow. You do not need to call these tools yourself. Describe what you want in plain language and the agent picks the right ones.

Once connected, your client lists the WeWeb tools available in the session. Many clients let you control how those tools are used, for example whether the agent can run a tool automatically or must ask for your approval first.

Tool permissions in Claude Desktop

This screenshot is from Claude Desktop, but other clients often offer similar controls in their own settings.

In Claude Desktop, open Settings → Connectors, then open your WeWeb connector. Under Tool permissions, choose for each tool or group of tools whether Claude can run it automatically, needs your approval, or is blocked.

See all available WeWeb tools →

Available tools

Once connected, your agent can call the tools below. They are grouped by area. You don't need to memorize them. Describe your goal in plain language and the agent picks the right tools.

Project & general

Tool	Description
listWorkspaces	List the WeWeb workspaces accessible to your account
switchWorkspace	Switch the active workspace for the session
listProjects	List the WeWeb projects available to your account
switchProject	Switch the active project for the session
getWeWebDocumentation	Look up WeWeb documentation to ground its work
getPageWorkflowTriggersContext	Get available triggers for page workflows
getAppWorkflowTriggersContext	Get available triggers for app workflows
getFrontendWorkflowActions	List available frontend workflow actions
getBackendWorkflowActions	List available backend workflow actions
wewebFeedback	Send feedback about the MCP experience
ping	Health check the connection

Pages & UI elements

Tool	Description
searchPages	Find pages in the project
createPage	Create a new page
updatePage	Update a page's settings
getElementsDocumentation	Get documentation for WeWeb elements
searchPageElements	Find elements on a page
getPageElementsVariables	List variables used by a page's elements
getElementLocalContext	Get the local context of an element
getElementParentStates	Inspect parent states of an element
getOtherPageSemantic	Read the semantic structure of another page
editElement	Edit an element's properties or style
editRepeat	Configure a repeat (list) on an element
addSection	Add a section to a page
reuseSection	Reuse an existing section (from another page) on the page
addElements	Add elements to the page
replaceSection	Replace an existing section
replaceElement	Replace an existing element

Design system

Tool	Description
searchTokens	Find design tokens
createTokens	Create design tokens (colors, typographies, spacings)
editToken	Edit a design token
deleteToken	Delete a design token
getGuidelines	Read the design system guidelines
editGuidelines	Edit the design system guidelines
searchClasses	Find CSS classes
createClass	Create a CSS class
editClass	Edit a CSS class
deleteClass	Delete a CSS class
activateIconSet	Activate an icon set

Variables

Tool	Description
searchVariables	Find variables in the project
createVariable	Create a variable
editVariable	Edit a variable
deleteVariable	Delete a variable

Frontend workflows

Tool	Description
searchGlobalWorkflows	Find global (app, page, function) workflows
createOrEditElementWorkflow	Create or edit an element-level workflow
deleteElementWorkflow	Delete an element workflow
createAppWorkflow	Create an app workflow
editAppWorkflow	Edit an app workflow
createPageWorkflow	Create a page workflow
editPageWorkflow	Edit a page workflow
createFunctionWorkflow	Create a function workflow
editFunctionWorkflow	Edit a function workflow
deleteGlobalWorkflow	Delete an app, page, or function workflow

Tables & data (WeWeb backend)

Tool	Description
listAllTables	List all tables
createTable	Create a table
dropTable	Delete a table
renameTable	Rename a table
addColumn / dropColumn / renameColumn / modifyColumnType	Manage table columns
addConstraint / dropConstraint	Manage table constraints
getTableRows	Read rows from a table
addRow / updateRow / deleteRow	Manage table rows
createTableIntegration / deleteTableIntegration / renameTableIntegration	Manage table integrations
listStorageFiles	List files in storage

Table views (interface data)

Tool	Description
searchTableViews	Find table views
fetchTableViewData	Fetch data from a table view
fetchRawTableData	Fetch raw table data
createTableView / updateTableView / deleteTableView	Manage table views
createTableViewIntegration / updateTableViewIntegration	Manage table view integrations

Authentication & users

Tool	Description
installAuthProvider	Install an authentication provider
listInstalledAuth	List installed auth providers
editAuthProviderConfig	Edit an auth provider's configuration
deleteAuthProviderConfig	Remove an auth provider
listUsers / createUser	Manage users
listRoles / createRole / deleteRole	Manage roles

Backend workflows & environment

Tool	Description
listAllBackendWorkflows	List backend workflows
getBackendWorkflow	Read a backend workflow
createBackendWorkflow / editBackendWorkflow / deleteBackendWorkflow	Manage backend workflows
listAllEnvVariables	List environment variables
createEnvVariable / editEnvVariable / deleteEnvVariable	Manage environment variables
getWeWebBackendStatus	Check the WeWeb backend status
installWeWebBackendUrl	Connect a WeWeb backend to the project

Tips

• Select a workspace and project first. Ask the agent to run listWorkspaces and switchWorkspace, then listProjects and switchProject, before making changes so it acts on the right project.
• Set up backend work before interface work. When a request spans both, have the agent create tables and backend workflows before building the UI that depends on them.
• Review as you go. Because tools run against your live project, keep an eye on changes and use the WeWeb editor and Preview to verify results.

Workaround for plugin actions

The MCP cannot create or edit plugin actions directly. Prefer Integrations when they cover your provider (for example Xano or Supabase), because those actions are available to the AI through MCP.

If you are stuck on plugins for now, some users use this makeshift workaround. It is not an official MCP feature. Treat it as a temporary bridge, and review every workflow carefully before you rely on it.

How the workaround works

In WeWeb, a workflow is stored as JSON. When you copy a workflow in the editor, WeWeb puts that JSON on your clipboard, which allows you to paste it as a new workflow in the WeWeb editor.

However, if you try to paste it into a text field, then you will see the raw JSON (which opens up the potential for you to make direct edits to the JSON).

If you already have workflows that use the plugin actions you care about, you can give that JSON to your MCP-connected AI as examples. The AI can study how those plugin actions are shaped, then return a full updated workflow JSON for you to paste back into WeWeb.

This only works well when you provide solid examples first. Without a few manually configured plugin-action workflows as references, the AI is guessing.

Steps

1. In WeWeb, manually build one or more workflows that already use the plugin actions you need (for example a Xano plugin action you configured yourself). Keep a few variations if you can. The more examples you give the AI, the better.
2. Open one of those workflows. In the top right of the workflow panel, open the ... menu next to the workflow name, then click Copy workflow.
3. Paste into your AI chat (ChatGPT, Claude, Cursor, and so on). You should see the workflow JSON.
4. Repeat for other example workflows if you have them, or keep them in the same chat as reference material.
5. Ask the AI to create or update a workflow based on those examples. Be specific about the change you want (for example which action to add, which fields to bind, or which conditions to set).
6. Ask the AI to return the full workflow JSON in the same format you pasted in. Do not ask it to summarize or partially rewrite the JSON.
7. Copy the JSON the AI returns.
8. In WeWeb, paste it as a new workflow. Open the result and check that the plugin actions, bindings, and flow look right.

What to expect

• The AI is not calling plugin tools through MCP here. You are doing a copy, edit, and paste loop with workflow JSON.
• Small mistakes in the JSON can produce a broken or incomplete workflow. Always open the pasted workflow in the editor and test it.
• This does not replace Integrations. When an Integration exists for your provider, use that instead so the AI can work through MCP tools normally.

Related

• Learn more about WeWeb AI →
• Learn more about AI modes →
• Learn more about passing context to AI →
• AI CRM with MCP build guide →