Agentica — разработка Python-агентов с MCP

ai-tooling

agentica-sdk is a Claude Code skill that guides building Python agents with the Agentica framework (v0.3.1), covering the @agentic decorator, spawn(), stateful persistence, multi-agent orchestration, and MCP tool integration. The skill documents core patterns including typed return values (int, dict, None for side-effects), per-call and decorator-level scope for passing tools, and the distinction between asynchronous spawn() and synchronous Agent() for use inside __init__. Model selection spans OpenAI (gpt-4.1, gpt-5), Anthropic (claude-sonnet-4.5, claude-opus-4.5), and any OpenRouter slug, with MaxTokens offering fine-grained control over per-round and per-invocation limits. Developers automating workflows with LLM agents, integrating third-party SDKs like Slack or Tavily via MCP config files, or debugging agent token usage will find this skill directly applicable.