Swain API Integration connector

Swain MCP connector

Build reliable Swain API integrations with your coding agent

Swain API Integration helps coding agents build against APIs generated and deployed by Swain. Select an authorised connection, inspect the deployed API schema, choose safe integration recipes, plan CrudSQL requests, and validate the exact request artifact before changing application code.

Connect

  • MCP endpoint: https://api.swain.technology/mcp
  • Transport: Streamable HTTP
  • Authentication: OAuth 2.0 authorisation code with S256 PKCE
  • Default scope: connections.read

You need a Swain account with access to at least one connection. Each user authorises the connector with their own Swain identity. The connector inherits that user’s tenant and connection permissions.

What the default connector can do

The default directory connection is read-only. It can:

  • list and select connection API targets the user can access;
  • check endpoint, authentication, tenant, schema, and environment readiness;
  • inspect the deployed OpenAPI schema and model guidance;
  • return integration recipes and safe request plans;
  • validate and explain CrudSQL request artifacts; and
  • provide a seeded demo target when a real connection is not ready.

The connector is a development control plane. It does not execute application data calls against the generated customer API. Application code produced from the plan calls the selected generated API directly using runtime credentials.

Swain also has separately authorised administrative tools for generated-auth provisioning and deployment. Those tools require explicit management scopes, are not granted by the default directory authorisation, and use confirmation-gated plan/apply workflows.

Example prompts

  1. Use Swain to inspect my selected connection and plan a tenant-scoped TypeScript read for active customers. Validate the request before editing code.
  2. Explore the Order schema and show me the safe way to upsert an order with its line items. Keep authentication and tenant values as runtime placeholders.
  3. Check the selected Swain API health, then repair this CrudSQL filter so it uses valid fields and relationships.
  4. Use the seeded Swain demo target to show the complete plan-and-validate workflow without changing production data.

Data handling

The connector processes the account, tenant, connection, deployed API schema, and request-planning inputs needed to answer the user’s request. It does not request Claude memory, unrelated conversation history, or arbitrary uploaded files.

Operational telemetry records the tool name, workflow phase, outcome, duration, input field names, error category, and server-recommended next tool. It does not record OAuth bearer tokens, API keys, database credentials, full tool argument values, generated request bodies, or customer API response data in MCP journey events.

For details about collection, use, retention, subprocessors, and user rights, see the Swain privacy policy and service-provider list.

Troubleshooting

  • If authorisation fails, disconnect and reconnect the connector so it can complete OAuth discovery and PKCE again.
  • If the token covers multiple tenants, select the intended Swain tenant during authorisation.
  • If no connection is available, create or obtain access to one in Swain, or start with the seeded demo target.
  • If health or schema checks return a blocker, follow the response’s nextTool, nextArguments, and blockingUntilFixed fields before planning again.