dsh-openapi

When AI agents interact with external APIs, directly loading large OpenAPI specifications and invoking endpoints can introduce security risks, context overload, and validation challenges. The dsh-openapi plugin addresses these pain points by providing a structured, secure gateway for DeepSeek Harness, enabling AI agents to efficiently explore and operate on OpenAPI 3.x APIs. Its core capabilities are delivered through three model-facing tools: openapi_list for discovering and searching API operations, openapi_describe for retrieving parameters, request bodies, and response details, and openapi_call for executing validated invocations—all augmented with built-in controls like parameter validation, credential mapping, SSRF protection, and response truncation. These features ensure API access is both flexible and rigorous, without burdening the model context with entire specifications.

This plugin is ideal for developers, AI engineers, and security-conscious teams building applications that require AI agents to call external services in a controlled manner. It shines in scenarios involving sensitive data, fine-grained access control (such as restricting to read-only methods), or deployment in complex network environments. By combining configured API catalogs with environment-based credentials, dsh-openapi simplifies integration while strengthening security boundaries, making API calls from AI agents more predictable and manageable.