Tried `Alishahryar1/free-claude-code` Today: One Terminal, Multiple Coding Agents

Описание и применение

Унифицированный терминальный подход к работе с несколькими кодающими агентами, позволяющий экономить токены и гибко переключаться между инструментами.

Задача

Developers need a flexible, cost‑efficient way to experiment with multiple coding agents without wiring each workflow to separate paid APIs.

Результат

A unified, cost‑effective setup that allows developers to experiment with multiple coding agents from a single terminal or IDE, benefiting from prompt caching and flexible tool integration.

Рабочий процесс

• 1. Set environment variables for B‑Lost endpoint, API key, and model (e.g., claude-fable-5).
• 2. Export these variables in your shell profile for persistence.
• 3. Use any compatible client (Cursor, Cline, Roo Code, Windsurf, Aider, etc.) to send prompts to the endpoint.
• 4. Leverage native Anthropic /v1/messages prompt caching to reduce costs on repeated repository context.
• 5. Switch between terminal, IDE, or mobile clients without reconfiguring the workflow.

Доказательства

The endpoint also supports native Anthropic /v1/messages prompt caching. For repeated repository context, that matters: cache hits receive a 90% discount, so the cost profile improves as the agent reuses the same system instructions, project files, and coding conventions.

Источник