How does a developer maintain conversation context across multiple API calls with Claude?
Select an answer to reveal the explanation.
![Multi-turn conversation state management infographic. Turn 1: [user: hello] → response. Turn 2: messages array = [user: hello, assistant: hi, user: next question] → send full history. Token budget growing indicator. Code pattern showing history accumulation. Stateless API reminder. 1024x768.](/media/anthropic/ccao-f/CCAO-F-e1-q073.webp)
Short Explanation
Every new API call starts fresh. To keep context, you append each exchange to the messages array and send the entire growing conversation with every call.
Full Explanation
Because the Claude API is stateless, maintaining conversation context requires the developer to include the full conversation history in each API call's messages array. The pattern is: start with user message → get assistant response → append both to history → next user message → append → send full history as messages array → get next response. This approach gives developers full control over context management. The option about using a 'session_id' parameter to link API calls automatically is wrong because no session_id parameter exists in the Anthropic API for stateful sessions. The option saying Claude automatically stores conversation history server-side, accessible via a conversation endpoint, is wrong because no server-side conversation storage exists. The option about using cookies to pass conversation context between API calls is wrong because cookies are a web concept unrelated to this API design.