Monte

Monte is an AI developer tool that acts as a portable persona layer for AI agents, giving them task-specific context before work begins through a CLI and API. Instead of pasting the same preferences into every chat, developers install the CLI (npm i -g monte-engine), authenticate, and run a personalize command that returns a task-aware instruction block agents can consume. Monte captures how you think, decide, communicate, handle risk, and evaluate outputs as a signal you own, then delivers it to tools like Claude Code, Codex, Notion, and other agents so each session starts aligned. The persona model is built on nine dimensions such as risk tolerance, time preference, learning style, and decision speed, drawn from sources, memories, and rules rather than open-ended memory. Output surfaces include a CLI command, an API endpoint (/personalization/context), and a portable prompt block that pastes directly into an agent session. Monte positions itself as fit rather than memory: durable context that reduces cold starts, cuts repeated preambles, and produces cleaner handoffs between tools. It works with hosted bearer keys and is aimed at developers who run multiple AI agents and want consistent, personalized guidance across all of them.

Developers use Monte to give their AI agents a portable, task-specific persona so every tool starts aligned with how they work.