Configure built-in actions, connect third-party integrations, register custom tools via OpenAPI, and control which tools your LLM can call autonomously.
Built-in actions
Cotonity ships with a library of built-in actions covering the most common agent tasks: HTTP requests (GET, POST, PUT, PATCH, DELETE), database reads and writes, email and SMS sending, file storage operations, and data transformation utilities like JSON parsing, string formatting, and array filtering. Built-in actions are pre-authenticated within the platform and require no additional credential setup. Find them in the node picker under the Actions category. Each built-in action has a schema that defines its required and optional inputs, and all output fields are documented in the platform's inline help panel.
Connecting external integrations
For actions beyond the built-ins, Cotonity integrates with hundreds of third-party services through its connector library — including Salesforce, HubSpot, Slack, GitHub, Google Workspace, Stripe, and more. To use a connector, navigate to Settings > Integrations, find the service, and click Connect. You will be redirected through the service's OAuth flow or prompted to enter an API key. Once connected, the service's actions appear in the node picker under the Integrations category. Credentials are stored encrypted and are scoped to your workspace; individual agents cannot access credentials from other workspaces.
Custom tools via OpenAPI schemas
If the service you need is not in the connector library, you can register any HTTP API as a custom tool by uploading its OpenAPI 3.x schema. Cotonity parses the schema, generates a node for each operation, and makes those nodes available in the node picker under Custom Tools. You supply authentication details (bearer token, API key header, or OAuth2 client credentials) when registering the tool. Custom tool nodes behave identically to built-in action nodes: they accept mapped inputs, execute the HTTP call, and expose the response fields to downstream nodes.
Granting tool access to the LLM
When you include an LLM node in your agent, you can optionally expose a subset of your configured actions as tools the LLM can call autonomously. In the LLM node's configuration, toggle Enable tool use and select which actions to expose. The LLM will receive a description of each tool (auto-generated from the action's schema, or overridden by you) and may decide to call one or more tools to fulfill the current step's goal. Restrict tool access to only the actions relevant to the current step to keep the agent's behavior predictable and to minimize unnecessary API calls.