Kimi-Researcher

Score 8.0

Moonshot AI's end-to-end deep research agent with autonomous multi-step search.

Built by: Moonshot AI · China startup
Categories: General agents
Pricing: Freemium（Limited free access during beta; membership entitlements per the official site.）
Open source: No
Self-hostable: No
International version: Yes
Data leaves China: Optional

"Data leaves China" means whether your data may be transferred outside mainland China when using the product.

Overview

Kimi-Researcher is a deep research agent released by Moonshot AI in mid-2025, trained end-to-end with reinforcement learning. Give it an open-ended question and it plans autonomously, runs many rounds of search (about 23 tool calls per task on average), cross-checks sources, and returns a cited report. It lives inside kimi.com and fits research-shaped work — industry studies, competitive analysis, literature reviews — where you hand over a question and expect a report. Its benchmark results were top-tier at release. Unlike Manus-style general executors, Researcher focuses on research deliverables rather than operating a computer; unlike classic search-augmented chat, it is an order of magnitude more autonomous and verification-heavy. Good fit for analysts, researchers, and consulting and investment teams. Moonshot also ships an open-source Kimi CLI for developers.