Topic collection

AI API Developer Tools

Integration guides for MCP, command-line tools, agents, and unified model API workflows.

These implementation guides are for developers connecting multiple AI models through one API. Topics include MCP, command-line workflows, agent integrations, authentication, task polling, error handling, and model discovery. Each guide is intended to shorten the path from an API key to a working request while keeping model-specific behavior visible. The request shape is the same for every media model: POST /v1/tasks with a model id and an input object, then poll GET /v1/tasks/{id} or pass a callback URL to be notified when the result is ready. Adding dry_run: true returns the exact price for those settings without running the task, which is the safe way to test a configuration from code or from an agent. Text models use the OpenAI request format at /v1/chat/completions and the Anthropic format at /v1/messages, with streaming and tool use, so existing SDK code usually needs only a new base URL and key. Failed generations are never charged. Begin with the interface that matches your environment (MCP for agents, the CLI for scripts and terminals, plain HTTP for services), then use the docs and the examples repository for production handling: retries, timeouts, webhooks, and polling intervals.

2 articles