Resources · Claude connector

Claude Lovable connectorWhat Claude can do in your Lovable account.

The Claude Lovable connector exposes 35 tools. 18 read your workspaces and projects, 11 build, publish or change something, and 6 are documented nowhere. Below: what Claude can ship from a chat, what spends credits, what it asks first and who can switch it off.

Verified Trustpilot reviews · AI, automation & growth agency

Overview

What changes once Claude can drive Lovable

Lovable builds web apps from plain-language prompts. With the connector, Claude talks to Lovable for you: it can start a project from your description, pass your follow-up requests to Lovable's own agent, read the resulting code and publish the app. You stay in one conversation instead of juggling the Lovable editor and a separate chat.

Go from idea to running app. create_project starts a project from your brief, and send_message relays each change you ask for to the Lovable agent, then waits for its answer.

See exactly what changed. get_diff, list_edits and read_file let Claude walk you through the code an edit touched, without opening the editor.

Ship and watch. deploy_project publishes the app and returns its live address, and get_project_analytics brings back visits, page views and traffic sources once it is out.

Where it stops: six listed tools have no description in any Lovable source, so this page says nothing about them. Publishing is refused while critical security findings remain open. Lovable says tool calls run live on your real projects, and building or messaging spends credits. Nothing starts by itself: Claude acts only while answering you, never on a schedule or after an event. For other app connectors, see our Integrations hub.

Vocabulary

Five words before you connect

The terms around this connector, kept short.

Connector
A link you set up once between Claude and an account you already have, so Claude can work in it while it answers you.
Tool
One named capability the connector gives Claude. It picks which to call mid-answer, and the directory lists each one by name.
Authorization
The service's own login screen, where you grant the access Claude will use. Granted once per person, and you can take it back.
Approval
The confirmation Claude waits for before it changes something in an account, shown in the conversation when it matters.
MCP
The shared standard connectors are built on, the piece that lets an assistant like Claude talk to an outside service.
Connect

Connect Lovable to Claude in three steps

  1. 01

    Find Lovable in Claude

    Open Claude's settings, go to Customize then Connectors, and look for Lovable. On a Team or Enterprise workspace, an Owner or Primary Owner enables the connector before members can each sign in.

  2. 02

    Log in to Lovable

    Click Connect on its row, then log in to Lovable in the window it opens. If the link breaks later, Disconnect and connect again. Lovable also describes a route from the chat itself: the + sign, then Connectors.

  3. 03

    Read the authorization screen

    Go through Lovable's authorization screen before accepting. It is Lovable's screen, not Claude's, and it sets what the access covers. Approval before each action remains Claude's default afterwards.

Tools

The 35 tools: 18 read, 11 write, 6 undocumented

Lovable gives Claude 35 tools: 18 that read your account, 11 that change something in it, and 6 no official source describes.

Names stay exactly as Claude shows them. The split comes from Lovable's tool reference, not from the names.

  • 18 read
  • 11 write
  • 6 not documented

What Claude reads (18)

18 tools

Tools that look at your account, projects, code, knowledge and analytics without altering them.

get_me

Returns the profile of the signed-in Lovable user along with the workspaces they belong to, which tells Claude whose account it is working in.

When it helps
you juggle a personal and a company login and want to confirm which one is connected.

Sourcedocs.lovable.dev · October 1, 2026 ↗

list_workspaces

Lists every workspace you belong to. Claude uses the result to pick the right workspace ID before looking for projects. get_me returns the same workspaces next to your profile.

When it helps
you belong to several workspaces and need to name the right one before a search.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_workspace

Shows a workspace's details, including its plan, credit balance, member count and settings. Claude reads it when you need the state of the workspace rather than of one project.

When it helps
before a long build session, you want to know how many credits remain in the workspace.

Sourcedocs.lovable.dev · October 1, 2026 ↗

list_projects

Searches and lists the projects of a workspace, with filters and full-text search, so Claude can locate one by name or by a keyword it contains.

When it helps
you remember a dashboard prototype from the spring but not what you called it.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_project

Returns a project's details: editor link, preview link and a screenshot of its current state. The public address of a published app comes from deploy_project instead.

When it helps
a teammate asks what the onboarding prototype looks like right now, before a meeting.

Sourcedocs.lovable.dev · October 1, 2026 ↗

list_template_projects

Lists the project templates available to start from, so Claude can suggest one that fits your idea instead of a blank start.

When it helps
you want a booking app and would like to know whether a ready-made base exists.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_message

Fetches the status and content of a message exchanged with a project's agent, typically to check on a request Claude did not wait for.

When it helps
a large request to the agent is still running and you want to know where it stands first.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_diff

Returns the unified diff of a message or a commit, the line-by-line view of what the code gained and lost. It pairs with list_edits, which pages through the project's history.

When it helps
after a design tweak, you want to be sure nothing outside the header moved.

Sourcedocs.lovable.dev · October 1, 2026 ↗

list_files

Lists all the files of a project at a given point of its code history, the map Claude needs before opening anything in particular.

When it helps
a freelancer asks how the project is organised before quoting a job on it.

Sourcedocs.lovable.dev · October 1, 2026 ↗

read_file

Reads the contents of a single file at a given point of the code history, so Claude can explain it or go through it line by line with you.

When it helps
you want the main App.tsx explained in plain words before asking Lovable for anything.

Sourcedocs.lovable.dev · October 1, 2026 ↗

list_edits

Lists a project's history page by page, the trail of everything that happened to its code over time. Claude reads it to tell you what came when.

When it helps
something broke yesterday and you want the latest three steps laid out to spot the culprit.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_workspace_knowledge

Reads the workspace knowledge, the custom AI instructions Lovable applies to every project in that workspace. That text is rewritten with set_workspace_knowledge, which replaces it in full.

When it helps
new projects keep picking a style you did not expect, and you suspect a workspace-wide instruction.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_project_knowledge

Reads the knowledge of one project, the custom AI instructions that apply to it alone and not to the rest of the workspace.

When it helps
you inherit a project and want to see the rules its builder gave Lovable before touching anything.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_database_status

Checks whether Lovable Cloud is enabled for a project, which tells Claude if there is a database to query at all.

When it helps
a database question returns nothing and you want to confirm Lovable Cloud is actually on for that project.

Sourcedocs.lovable.dev · October 1, 2026 ↗

list_connectors

Lists every connector integration available in a workspace, whether standard, built in or MCP-based, and shows which ones are switched on.

When it helps
you are planning a feature that needs payments and want to know what Lovable already offers for it.

Sourcedocs.lovable.dev · October 1, 2026 ↗

list_connections

Returns the connector accounts already authenticated in a workspace, as opposed to the ones that are merely available. Claude can tell you which outside accounts are really linked.

When it helps
an app feature fails and you want to rule out a missing account link.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_project_analytics

Pulls historical analytics for a published project: visits, page views, bounce rate, visit length, broken down by page, source, device and country. It needs a start and an end date.

When it helps
a month after launch, you want to know where visitors come from.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_project_analytics_trend

Gives the real-time visit count and the short-term trend of a published project, the live view next to the historical one.

When it helps
you just shared the link in a newsletter and want to see if people are arriving.

Sourcedocs.lovable.dev · October 1, 2026 ↗

What Claude builds or changes (11)

11 tools

Tools that create, publish, overwrite or open something in your Lovable account. No source describes their confirmation, so the general rule applies.

create_project

Approval: see the rule

Starts a new project from a build prompt handed to Lovable's agent, optionally from a template or a design system. It spends standard Lovable credits.

What Claude asks for
no source describes a confirmation for this tool; the general rule below applies.
When it helps
you describe a feedback form app in two sentences and want a first build.

Sourcedocs.lovable.dev · October 1, 2026 ↗

deploy_project

Approval: see the rule

Publishes a project to Lovable's hosting and returns the live address. Lovable says who can reach that address depends on your Lovable plan and workspace settings.

What Claude asks for
nothing tool-specific is documented; lean on the general rule.
When it helps
the prototype is ready for a client demo tomorrow.
Watch out
publishing is refused while unresolved critical security findings remain.

Sourcedocs.lovable.dev · October 1, 2026 ↗

remix_project

Approval: see the rule

Copies an existing project into a workspace. The source must be reachable from it: your own projects, templates, or projects open to public remixing.

What Claude asks for
no confirmation is on record here; the default rule covers it.
When it helps
you want to try a bold redesign without touching the version a client already uses.

Sourcedocs.lovable.dev · October 1, 2026 ↗

set_project_visibility

Approval: see the rule

Changes who can access a project in the editor, matching the options of the share dialog. It governs editor access only, not who can visit the published app.

What Claude asks for
the sources stay silent on a confirmation; see the general rule.
When it helps
a contractor joins, and the project should be view-only for the wider workspace.

Sourcedocs.lovable.dev · October 1, 2026 ↗

set_folder_visibility

Approval: see the rule

Switches a folder between personal and workspace visibility, and the change carries over to every project inside it.

What Claude asks for
no specific confirmation is described; the default rule applies.
When it helps
a set of drafts is ready for the team to see at once.

Sourcedocs.lovable.dev · October 1, 2026 ↗

send_message

Approval: see the rule

Passes a request to a project's own AI agent and waits for the reply. It can ask for a plan first, attach files, or return at once and be checked later. Each call spends credits.

What Claude asks for
nothing documented beyond the general approval rule.
When it helps
adding a dark mode toggle without opening the editor.

Sourcedocs.lovable.dev · October 1, 2026 ↗

set_workspace_knowledge

Approval: see the rule

Writes the workspace knowledge, the instructions that apply to every project, and replaces the existing text entirely rather than adding to it.

What Claude asks for
no source describes a confirmation; the general rule governs it.
When it helps
your brand guidelines changed and every future build should follow them.
Watch out
keep a copy of the old text, since it is overwritten.

Sourcedocs.lovable.dev · October 1, 2026 ↗

set_project_knowledge

Approval: see the rule

Writes the instructions of a single project, replacing whatever was there before in full.

What Claude asks for
nothing specific is documented; the default rule stands.
When it helps
one client app must always use their accessibility rules, unlike the rest of the workspace.
Watch out
the previous instructions are not kept.

Sourcedocs.lovable.dev · October 1, 2026 ↗

enable_database

Approval: see the rule

Turns on Lovable Cloud for a project, the condition for query_database to have a database to query. Lovable says activation takes 30 to 60 seconds.

What Claude asks for
no tool-level confirmation is described; see the general rule.
When it helps
your prototype now needs to store sign-ups instead of losing them.

Sourcedocs.lovable.dev · October 1, 2026 ↗

query_database

Approval: see the rule

Executes an SQL query on a project's Lovable Cloud database. Lovable is explicit that it covers reads, writes and schema changes, with your full database permissions.

What Claude asks for
no source describes a confirmation; the general rule is the reference.
When it helps
counting this week's sign-ups straight from the table.
Watch out
the same tool can alter or drop data.

Sourcedocs.lovable.dev · October 1, 2026 ↗

get_file_upload_url

Approval: see the rule

Returns a presigned address to upload a file that will travel as an attachment with a message to the Lovable agent.

What Claude asks for
nothing on record beyond the general approval rule.
When it helps
you want Lovable to rebuild a screen from a mockup image sitting on your laptop.

Sourcedocs.lovable.dev · October 1, 2026 ↗

Undocumented (6)

6 tools

Listed on the directory sheet, absent from Lovable's documentation. We do not guess what they do from their names.

list_library_projects

This name appears on the directory sheet only. Lovable's MCP page, read on 1 October 2026, does not list it, and neither does the machine-readable tool reference Lovable says matches its server.

Watch out
the general approval rule below covers it like any other tool.

get_database_connection_info

Same situation: listed in the directory, missing from both Lovable references we read. We will not infer its purpose from its wording, so the page stops at that.

Watch out
whatever it returns, it falls under the default approval rule described further down.

list_mcp_servers

No Lovable source we found mentions this tool. Lovable's documentation instead names connector tools that are absent from the sheet. Neither Lovable's MCP page nor its machine-readable reference, both read on 1 October 2026, describes it.

Watch out
the general approval rule still applies to it, as to every listed tool.

list_mcp_catalog

The directory publishes the name and nothing more. Neither the Lovable MCP page nor its tool reference describes it, so any description here would be a guess.

Watch out
it sits under the same approval rule as the rest of the list.

add_mcp_server

Listed on the directory sheet, undocumented by Lovable as of 1 October 2026. This page gives it no capability at all, because no official text grants it one.

Watch out
treat it with the general approval rule; nothing more specific exists for it.

remove_mcp_server

One more name the sheet shows and Lovable's references leave out. Until Lovable documents it somewhere public, this page has nothing reliable to add about it.

Watch out
the default confirmation rule below is the only framework that covers it.
Approvals

What Claude checks with you first

By default, Claude stops and asks for confirmation before each action it takes on an account in someone's place.

No source describes a confirmation for any of the eleven Lovable tools that build, publish, overwrite instructions, open a database or run SQL, so the default rule covers them all. That matters here: Lovable says calls run live, edit real projects and spend real credits. On Team and Enterprise, workspace owners decide whether members may let some actions through without a prompt, and they can restrict the connector for everyone. Lovable adds that the server inherits your Lovable permissions exactly, with nothing extra.

Plans

Which plans this works on

Plan availability is published for no single connector: 0 of the 819 directory sheets shows it.

The published rule is general. Remote connectors are open to all users on Claude, Cowork, Claude Desktop and mobile, and on Team and Enterprise an Owner or Primary Owner enables the connector before members connect. Lovable's documentation answers the question for its own plans, not Claude's; the limits section puts both side by side. For the live state on your account, check the Lovable sheet in the official directory.

Limits

Where this connector stops

A connector is not an automation. Claude calls Lovable while it answers you; a new visit, a failed build or a teammate's edit does not start anything on its own.

Only Lovable documents this connector; no Claude help page covers it. Six tools on the sheet are missing from Lovable's references, while those references name other tools the sheet lacks. The partner badge is not a security audit, and Anthropic says on every sheet that it does not choose a publisher's tools or guarantee their behavior. Lovable also mentions a tool that imports a Claude design, available in Claude clients only and absent from the sheet.

Two official sources disagree

Which plans can use the Lovable connector?

What this page followsThey do not talk about the same plans: Lovable opens its server to every Lovable plan, Anthropic states the rule on Claude's side for all connectors. Neither is outdated. For Claude's side, check the directory sheet on your own account.

Need help

Need help connecting Lovable to Claude?

A person reads every message.

FAQ

Questions about the Claude Lovable connector

01What can Claude do with the Lovable connector?
With Lovable, Claude builds and manages apps from the conversation. It starts a project from a prompt or a template, pass change requests to Lovable's agent, read the code, diffs and edit history, publish the app and read its visit analytics. It can also read and overwrite workspace or project instructions, enable a database and run SQL on it. Six listed tools have no description in any Lovable source. Building and messaging the agent use Lovable credits.
02Can Claude publish or change my Lovable projects?
Yes. Eleven tools change something: creating or remixing a project, messaging the Lovable agent, publishing, changing project or folder visibility, overwriting instructions, enabling the database, running SQL that can read, write or alter the schema, and preparing a file upload. Lovable stresses that these calls hit real projects and spend real credits. Eighteen tools only read, and six remain undocumented, so nobody can say from the sources what they do.
03Does Claude ask before it acts in Lovable?
By default, Claude asks for confirmation before each action it takes on an account for someone. No source describes a confirmation specific to any Lovable tool, so that default rule covers publishing, messaging the agent, overwriting instructions and running SQL. On Team and Enterprise workspaces, owners decide whether members may skip the prompt for some actions, and they can keep reading open while blocking every change for the organization. Lovable stresses that calls run live, on real projects.
04Which plans is Lovable available on?
No official source publishes plan availability connector by connector, and none of the 819 directory sheets shows it. The general rule opens remote connectors to all users on Claude, Cowork, Claude Desktop and mobile, with an Owner or Primary Owner enabling them first on Team and Enterprise. Lovable says its server works on all of its own plans, which is a different question, quoted in the limits section. The directory sheet shows Claude's side.
05Does Claude see every project in my Lovable account?
Yes, within your rights. Lovable warns that the access covers your full account, not one project: Claude can list, read and edit every project you can reach. It inherits your Lovable permissions exactly and nothing more, so a project shared with a colleague but not with you stays out of reach. On a Team or Enterprise workspace in Claude, an owner can still narrow what the connector may do. Lovable describes no extra surface beyond that.
06Why does Claude say Lovable refused to publish my app?
Because Lovable will not publish a project while critical security findings stay unresolved. Lovable documents this check for deploy_project: as long as those findings stay open, the tool refuses to put the app online. The documentation does not detail how to handle them; that is settled inside Lovable, before you ask Claude to publish again. No setting on the Claude side bypasses that check: the block comes from Lovable itself.
07Should I use Claude or an automation tool with Lovable?
They do different jobs, so choose by use rather than by ranking. Claude with Lovable fits building you steer in conversation: describe, review the diff, adjust, publish. An automation tool fits steps that must run without anyone asking, such as reacting to an event on a schedule. The Lovable connector does not do that by itself, since Claude calls its tools only while answering you, and each of those calls runs live on your real Lovable account.