chat-ui

dbdaabac - Prompt LLM to use parallel tool calls for independent operations

Prompt LLM to use parallel tool calls for independent operations The MCP backend already executes multiple tool calls in parallel (Promise.allSettled), but the system prompt never told the model it could issue multiple calls at once. LLMs default to sequential tool use due to training biases. Adding an explicit PARALLEL TOOL CALLS instruction encourages models to batch independent tool calls in a single response, reducing round-trips and improving latency, while warning against parallelizing dependent calls. https://claude.ai/code/session_01GGYJ1UyJvkQL38oU7hNedP