Skip to main content
You can create your own connector for any REST API, whether it’s a public third-party API or an internal service, and make it available to your Lovable workspace. You define the connector once (its API endpoint, authentication method, and optional knowledge files), and everyone in your workspace can then connect to it like any other app + chat connector. Connectors you create are app + chat connectors: the apps you build use them to call the API, and Lovable can work with them while you build, guided by the knowledge files you add. To only bring context from your tools into the Lovable chat, use a custom MCP server as a chat connector instead.

Who can create connectors

Creating and managing connectors requires the workspace admin or owner role, and is available on all plans. Connectors you create are scoped to a single workspace: they appear in that workspace’s connector catalog and are not visible to other workspaces. Once created, workspace members connect to your connector the same way they connect to any other app + chat connector.

How to create a connector

You define the whole connector in a single form, opened from the connector catalog:
1

Open the connector catalog

Go to Connectors in your workspace. In the connector catalog, select the Custom card (“Connect your own API”). This opens the Create custom connector form.
2

Describe your connector

Fill in the basic details:
  • Display name: the name users see in the catalog. Use the name of the API the connector reaches (for example, Acme Projects API). The connector’s internal ID is derived from this name and can’t be changed later.
  • Short description: a one-line summary (for example, Project management boards).
  • Description: a longer explanation of what the connector does.
  • Category: where the connector appears in the catalog (Productivity, Sales, Marketing, E-commerce, Support, Messaging, Development, Payments, or Other).
  • Documentation URL (optional): a link to the API’s documentation.
Under Created by, enter the provider Name and Homepage URL. These are shown to users so they know who is behind the API.
3

Configure the API endpoint

Under API endpoint, set:
  • API base URL: where requests are sent. Must be an https:// URL.
  • Method and Verification path: an endpoint Lovable calls to verify credentials when a user connects (for example, GET /v1/me). The path must start with /.
4

Choose an authentication method

Under Authentication, pick how the API expects credentials (see Authentication methods below). For non-OAuth methods, a live preview shows what the agent’s requests will look like with your configuration applied.Credential values are never entered in this form. Users provide their own API key or OAuth client credentials when they create a connection.
5

Add knowledge (optional)

Under Knowledge, click Add knowledge file to teach Lovable how to use the API: key endpoints, request formats, pagination rules, or gotchas. Each knowledge file has a Name, Description, and Content. Good knowledge files dramatically improve how well Lovable uses your connector. See Writing connector knowledge files for guidance and an example.
6

Create the connector

Click Create. The connector now appears in your workspace’s connector catalog, and in the Custom connectors section of the connector admin settings.

Authentication methods

Pick the method that matches how your API expects credentials:

Setting up OAuth 2.0

For APIs that use OAuth 2.0, configure:
  • Authorization URL and Token URL: your provider’s OAuth endpoints (both must be https://).
  • Scopes (optional): one per line, or separated by spaces or commas.
  • Scope separator (optional): leave blank for the default (space). Some providers require a comma.
  • Use PKCE: enable if your provider supports or requires PKCE.
The form displays a Redirect URI to register with your provider. Copy it and add it to your OAuth app’s allowed redirect (callback) URLs. It must match exactly.

Writing connector knowledge files

Connector knowledge files are how Lovable learns to use your API. Without them, Lovable only knows your base URL and authentication method, so it has to guess endpoint paths, parameter names, and response shapes. A good knowledge file removes the guesswork, so treat it as documentation written for Lovable rather than for humans. They are separate from workspace and project knowledge: connector knowledge travels with the connector and only covers how to call its API. Each knowledge file has three parts:
  • Name: a short identifier (for example, Acme Projects API basics).
  • Description: a one-line summary of what the file covers, so Lovable picks the right file for the task (for example, Endpoints and request formats for reading and writing Acme Projects boards and tasks.).
  • Content: the actual instructions, in Markdown. This is where the endpoints, examples, and gotchas go.

What to include

Follow the same structure Lovable uses for its built-in connectors:
  1. A one-line summary of what the API does and what it’s for.
  2. Key endpoints: method, path, and what each does. Use paths relative to your API base URL (for example, GET /v1/tasks). Lovable routes requests and attaches credentials automatically, so don’t hardcode full URLs or authentication headers.
  3. Request examples for the most common operations, as code blocks. Show the exact body shape, required parameters, and query options.
  4. Response shapes: mention wrappers and field names Lovable should expect (for example, “responses are wrapped in a top-level data field”).
  5. A notes section for the things that trip up integrations: pagination parameters, rate limits, API versioning in paths, async operations, and date or ID formats.

What to leave out

A few things don’t belong in knowledge files:
  • Credentials or secrets: never put tokens, keys, or client secrets in knowledge content. Authentication is configured separately and injected automatically.
  • Authentication boilerplate: Lovable doesn’t need instructions for attaching credentials. The connector handles it.
  • Marketing copy: descriptions of how great the product is waste context. Stick to technical facts.

Example

A knowledge file for the fictional Acme Projects API, a project management API:
You can split knowledge across multiple files, for example one for core concepts and endpoints, and another for a complex sub-domain like reporting or webhooks. Each file’s Description helps Lovable pick the right one for the task at hand. A connector can have up to 50 knowledge files, and each file’s content can be up to 50,000 characters.

Using your connector

After creation, your connector behaves like any other app + chat connector:
  1. Users open Connectors, find your connector in the catalog, and click Connect.
  2. They enter the credential values you defined (an API key, username and password, or OAuth Client ID and secret). For OAuth connectors, a popup runs the authorization flow with your provider.
  3. They choose who can use the connection: only themselves (the default), specific people, or the whole workspace. See Who can use connections.
  4. Lovable verifies the credentials against the Verification path you configured.
  5. Once connected, the connection can be linked to projects, and Lovable can build apps that call your API, guided by any knowledge files you added.

Managing connectors

Workspace admins and owners manage the connectors created in their workspace from Connectors → Admin settings, under Custom connectors:
  • Availability: control who can create connections (No one, Admins, or Editors & admins), the same way as for any other connector. The per-connector picker is available on Business and Enterprise plans. On Pro, any workspace member with the editor role or higher can create connections. See Manage connectors in your workspace for the defaults on each plan.
  • Edit: update the connector’s details, endpoint, knowledge, and authentication configuration. Some authentication changes are restricted because existing connections store credentials under the current shape: a connector can’t switch between OAuth 2.0 and credential-based authentication, and existing credential field keys are locked. You can still switch between the single-credential methods (bearer token, custom header, or query parameter), or move any credential-based method to Advanced.
  • Delete: remove the connector definition from the workspace.
Deleting a connector removes its definition from the workspace, and existing connections for it stop working. This cannot be undone.

Security

Custom connectors follow the same security model as built-in connectors, described in Integration security:
  • Credential values are entered only when a connection is created, and are stored encrypted by Lovable. They are injected into requests server-side by the connector gateway, and are never exposed in chat, in project code, or to other users.
  • Workspace-shared connections can be used by other members of the workspace, so only share connections whose data you’re comfortable making available to your team.