AI article
Claude Agent SDK vs rolling your own tool loop: which one for your product?
Community description: Three ways to run a Claude agent: a hand-written loop on the Messages API, the Anthropic SDK's Tool Runner, and the Claude Agent SDK. Runnable TypeScript for each, measured token usage, and a decision table.
Dev.to | Nov 5, 2026 | Agentwright
Automated excerpt
Same API, same tools you define, but client.beta.messages.toolRunner() runs the loop. Claude Code packaged as a library (@anthropic-ai/claude-agent-sdk). The Tool Runner lives in the regular API SDK and only loops over tools you define.
Selected automatically from source text; not independently written or fact-checked. Read the original for full context.