Instinct FAQs

Question 1

What is Instinct and how does it benefit AI coding agents?

Accepted Answer

Instinct provides self-learning memory for AI coding agents. It observes your agent's behavior, patterns, and preferences, allowing it to remember, adapt, and make smarter suggestions without you needing to repeat instructions.

Question 2

Which AI agents and platforms are compatible with Instinct?

Accepted Answer

Instinct is compatible with any MCP (Model Context Protocol) enabled agent, including popular tools like Claude Code, Cursor, Windsurf, Goose, and Codex. It offers both CLI and Python library integration.

Question 3

How does Instinct learn and promote patterns into suggestions?

Accepted Answer

Instinct automatically records agent patterns, tracking confidence with each re-observation. Patterns are promoted from 'raw' to 'mature' (suggestions) at confidence >= 5, and to 'rule' (auto-apply) at confidence >= 10.

Question 4

What types of patterns can Instinct learn and remember?

Accepted Answer

Instinct can learn various patterns, including tool sequences (e.g., `seq:lint->fix`), user preferences (`pref:style=black`), recurring fixes (`fix:missing-import`), and effective tool combinations (`combo:pytest+coverage`).

Question 5

Does Instinct support learning across different coding projects?

Accepted Answer

Yes, Instinct uses project fingerprints to differentiate. It can store project-specific patterns that are suggested only in that context, and global patterns (without a project field) that apply across all your work, facilitating cross-project learning.

Instinct

Instinct

主な機能

ユースケース

主な機能

ユースケース