A kagi command-line tool to query the Kagi FastGPT search API (github.com)

A lightweight command-line client, kagi, lets you query the Kagi FastGPT search API from the terminal to get AI-generated answers grounded in web sources. The tool (installable via Homebrew, go install, or building from GitHub) accepts queries from arguments, stdin or pipes and returns clean, numbered answers with source references. Output can be plain text, Markdown or JSON (for automation), supports quiet/heading/debug modes, automatic terminal-aware color (with force/disable flags), configurable timeouts, and sensible exit codes for CI. You must set KAGI_API_KEY (or pass --api-key); the CLI surfaces common API/network errors and rate-limit responses. The repo is a minimal Go project (single main.go, 83 tests) under MPL-2.0 and is unofficial. For AI/ML practitioners the utility is practical for rapid research, reproducible documentation, and agent or CI workflows: JSON outputs and quiet mode make it easy to pipe answers into scripts, generate READMEs or feed downstream agents while retained web references improve traceability of claims. Technical implications include reliance on Kagi’s FastGPT for web-context grounding (and its rate limits), predictable behavior via timeouts/debugging, and simple integration points for toolchains — making kagi a handy, auditable bridge between research queries and programmatic workflows.