Crawl Foundry
API & MCP

REST API and MCP: start an integration

Use the REST API from your own software, or connect an MCP client through the application's installation flow. Both require authorized access to the correct workspace.

Before connecting an integration, set up your first workspace.

Read the current API reference

Swagger lists the supported endpoints, request fields and response schemas. The machine-readable OpenAPI document is the source for generating an API client. These references are public, but workspace requests require authentication.

Create and manage credentials in the application

1
Open the integration or agent settings inside the application and create a credential with only the permissions your integration needs.
2
Copy the key when it is shown and store it in your server's secret store or an environment variable. API keys are not sent by email. Do not put them in browser code or a public repository.
3
Revoke an unused or exposed credential in the application. Create a new credential when you need to replace it.

Start with a read-only request

With a credential permitted to list workspaces, request the accessible workspace list first. Replace the environment variable with your own securely stored key. Use the returned public workspace ID in subsequent workspace-scoped requests. This example does not start a crawl or buy fresh data.

curl --fail-with-body "https://app.crawlfoundry.com/api/v1/workspaces" \
  --header "Authorization: Bearer $CRAWL_FOUNDRY_API_KEY"

Review permissions and costs before running work

A working key does not authorize every action. Check the operation's requirements, permissions, workspace and available balance. Follow the API's estimate and authorization steps where required. For jobs, inspect status and errors rather than treating an accepted request as a completed result.

Before requesting fresh data, review how balance and data refreshes work.

Connect an MCP client

Use the client setup provided in the application for Claude, Cursor, VS Code or Codex. The customer endpoint is https://app.crawlfoundry.com/api/mcp. A workspace-bound connection includes the workspace selector generated by the app. Complete sign-in and consent in the browser. A connection without a fixed workspace still needs workspace selection for the session.

Begin by asking which workspace is active and requesting existing information. Approve any paid operation only after reviewing its scope and cost. Client-specific examples are being documented and tested separately.

Related product documentation