Lovable

Get the code diff from a message or commit. Pass message_id (from send_message) to see what the agent changed, or sha for a specific commit. Returns a unified diff.

Get a presigned URL to upload a file. Workflow: 1) call this to get upload_url and file_id, 2) PUT the file content to the upload_url, 3) pass the file_id in the files array of send_message or create_project.

Get the authenticated user's profile and first 100 accessible workspaces. When has_more is true, use list_workspaces to paginate through all workspaces. Without workspaces:read, returns the profile with an empty workspace list and instructions to enable workspace access.

Get the status and content of a message. Pass the message_id and thread_id from send_message. The top-level status field tracks the user message ('queued', 'accepted', or 'running') and never becomes terminal; read the agent's state from response.status: 'completed', 'stopped' and 'error' are terminal; 'awaiting_input' means the agent paused for a human decision and will not resume on its own. response is absent while the agent is still working. When response.awaiting_input is present the pause is a tool approval: show the user its tool_id and params and pass their decision to respond_to_approval. If awaiting_input.requires_secure_form is true, send the user to its editor_url instead and never collect a credential in chat. Without awaiting_input the pause is a check-in (such as the spend-limit prompt), answered in the Lovable editor. A new send_message supersedes a pause instead of answering it, unless it shows status 'queued' with queue_pause_reason 'hitl_tool': then it runs after the pause is answered.

Get details for a specific project. Returns editor_url (Lovable editor) and preview_url (static preview, always available after first prompt). Includes a screenshot of the app when the project status is completed. The top-level status is the provisioning status of the scaffold (in_progress, completed, failed), not the agent's build progress; project.agentFinished reports whether the latest agent turn has finished and is false while the agent is running or paused for awaiting_input.

Get historical analytics for a published project: visitors, pageviews, bounce rate, session duration, and breakdowns by page, source, device, and country.

Get real-time visitor trend for a published project. Returns the current visitor count and visit counts in 5-minute intervals over the last 30 minutes.

Get a project's knowledge content. Knowledge is custom instructions that guide the AI agent's behavior for this specific project.

Get details for a specific workspace: name, plan, project count, and the caller's role.

Get the workspace's knowledge content. Knowledge is custom instructions that guide the AI agent's behavior across all projects in the workspace.

Get a workspace skill, including its full SKILL.md contents. Use this before update_workspace_skill when preserving existing content.

List all connectors (integrations) available in a workspace. Connectors give the AI agent access to external services like Stripe, Supabase, Shopify, Slack, and more. Grouped by type: standard (OAuth-based API connectors), seamless (zero-config integrations), and MCP (Model Context Protocol servers). Each includes its enabled/disabled status. See https://docs.lovable.dev/integrations/introduction for details.

List connectors that have been added to a workspace (both from the catalog and custom). These are the connectors the AI agent can use during chat. To add a new one, call add_connector to get the dashboard URL the user should open.

List design systems available in a workspace. A design system is a reusable Lovable project that ships components, tokens, and styles. Pass any of these IDs as `design_systems` to `create_project` to start a new project from that design system.

List the edit history of a Lovable project. Each edit represents a code change with a commit_sha that can be used with get_diff. Supports cursor pagination via the cursor parameter (use pagination.next_cursor from the previous page).

List a page of files in a Lovable project. Optionally pass latest_commit_sha from get_project, or commit_sha from send_message/list_edits. Use pagination.next_cursor to fetch more pages.

List recent messages in a project, newest first. Use this to discover message IDs when you only have a project_id — for example, to fetch the latest assistant response with get_message or look up a diff with get_diff. Paginate older history by passing `pagination.next_cursor` as `cursor`.

Search and list projects in a workspace. Supports full-text search by name, AI-generated description, creator name, or email. Uses ngram matching for fuzzy search. Supports filtering by visibility, publish status, folder, and creator. Returns paginated project summaries with creation_status (in_progress, completed, failed). This describes scaffold setup; completed does not mean the agent has finished building the app.

List available template projects in a workspace. These can be used as template_project_id when creating a project to clone the template's codebase.

List workspace skills. Skills are reusable agent instructions stored in the workspace repo and available to projects in that workspace. Use this before updating or deleting a skill.

List all workspaces the user belongs to. Returns a concise summary; use get_workspace for full details. Call this first to discover workspace IDs, which are required by project creation, workspace settings, and connector management tools.

Move one or more projects into a workspace folder. Existing folder memberships are removed first. The folder may update project visibility to match folder visibility; the response lists added, skipped, and visibility-updated projects.

Execute a SQL query against a project's cloud PostgreSQL database. Returns rows as JSON objects. Supports SELECT, INSERT, UPDATE, DELETE, and DDL statements. The database must be enabled first (use get_database_status to check, then enable_database if needed). Caution: write operations (INSERT, UPDATE, DELETE, DROP) modify production data permanently — prefer SELECT for inspection unless a write is explicitly required.

Read the contents of a single file in a Lovable project. Optionally pass latest_commit_sha from get_project, or commit_sha from send_message/list_edits. Returns the raw file content as text.

Remix (fork) an existing project into a workspace. Creates a copy of the source project. Waits for the remix to complete and returns the new project details.

Send a chat message to a project's AI agent and wait for the agent to finish. The agent can write and edit code, install npm packages, create pages and components, set up authentication, connect integrations (Stripe, Supabase, etc.), and fix bugs. Write messages in natural language — describe what you want, not how to implement it. Attach images (designs, screenshots) to guide visual changes. Returns the agent's response, activity log, and any edits made. The preview_url is rebuilt once the agent finishes. Set wait=false to return immediately without waiting.

Set a folder's visibility. 'personal' restricts to the creator and collaborators. 'workspace' makes it visible to all workspace members. Changing visibility propagates to all projects in the folder. Nested folders cannot change visibility (move to root first). Personal folders require a Business or Enterprise plan.

Set a project's knowledge content. Knowledge is custom instructions that guide the AI agent's behavior for this specific project. Max 10,000 characters. Caution: this replaces existing knowledge entirely — read with get_project_knowledge first if preserving existing content. Examples of useful knowledge: coding conventions, preferred libraries, brand voice guidelines, API endpoint patterns, or 'always use Supabase for auth'.

Set who can open a project in the editor. 'workspace_edit' lets workspace members edit at their role's access. 'workspace_view' caps what workspace members get from their role at view, leaving explicit grants intact (requires Business plan or higher). 'restricted' limits access to the owner, invited collaborators, and the workspace owner (requires Business plan or higher).

Set the workspace's knowledge content. Knowledge is custom instructions that guide the AI agent's behavior across all projects in the workspace. Max 10,000 characters. Caution: this replaces existing knowledge entirely — read with get_workspace_knowledge first if preserving existing content. Examples of useful knowledge: coding conventions, preferred libraries, brand voice guidelines, API endpoint patterns, or 'always use Supabase for auth'.

Update a workspace skill by replacing skills/{skill_name}/SKILL.md in the workspace repo. Only workspace admins and owners can update skills. Read the existing skill first with get_workspace_skill if you need to preserve content.