Skip to main content
Agent Context System solves the cold-start problem for AI coding agents. Every session, your agent starts from zero — forgetting everything it learned about your project the day before. This system gives agents a shared, reviewable, version-controlled memory that persists across sessions and works across every major agent tool.

Quick Start

Get your agent reading persistent project context in under 5 minutes.

How It Works

Understand the two-file system and how knowledge flows from sessions to shared context.

Setup Guides

Step-by-step setup for GitHub Copilot, Claude Code, Cursor, and Windsurf.

CLI Reference

Commands for init, validate, and promoting patterns to shared context.

​
What You Get

Two markdown files. One committed, one gitignored. Your agent reads both at the start of every session and updates the local one at the end.

AGENTS.md

Your project’s source of truth. Committed and shared. Always loaded into every agent session. Under 120 lines of compressed project knowledge.

.agents.local.md

Your personal scratchpad. Gitignored. Grows over time as the agent logs what it learns each session — patterns, gotchas, dead ends.

Auto-Reflect

Agents observe significant events during sessions and surface patterns worth promoting to shared context at session end.

Auto-Consolidation

Daily logs compress automatically into curated topic files. 9:1 compression ratio, zero infrastructure required.

​
Supported Agents

Agent Context System works with every major AI coding tool:
1

GitHub Copilot

Install via the official Copilot skill registry with npx skills add.
2

Claude Code

Wired up automatically via CLAUDE.md symlink during init.
3

Cursor & Windsurf

Configured via .cursorrules and .windsurfrules pointing to AGENTS.md.
4

OpenClaw

Clone into your skills directory — reads AGENTS.md natively.
No plugins, no infrastructure, no background processes. The convention lives inside the files themselves, and the agent follows it.