Open Standard · v1.4

The open standard for
agent engineering.

Give your AI agents persistent memory, deep project knowledge, and real skills — structured once, useful forever, portable across every tool.

Get Started View on GitHub Try in 60s
.agent/
  • Identity
  • agent.yaml
  • instructions.md
  • permissions.yaml
  • Capabilities
  • skills/
  • workflows/
  • commands/
  • Persistence
  • memory/
  • registry/
The problem

Every session starts from zero

Your agent doesn’t know your architecture, conventions, or deployment. You re-explain every session.

The standard

Every session builds on the last

AES gives your agent structured memory, real skills, and deep project context. Checked into git, portable across tools.

Sync everywhere

One standard. Six tools.

Write once in .agent/ — sync to any tool with a single command.

Claude
Cursor
Copilot
Windsurf
OpenClaw
Codex
Get started

Choose your path

Whether you’re starting fresh or adding agent context to an existing project, AES gets you there in minutes.

Starting a new project

Pick a domain, scaffold .agent/, and start building with a capable agent from day one.

$ aes init
$ aes sync -t claude
  • Choose your mode — Dev-Assist or Agent-Integrated
  • Get pre-built skills and memory templates for your domain
  • Sync to your AI tool and start coding immediately
Learn how it works →

Organize an existing project

Run aes init in your repo to layer structured agent context on top of your existing codebase.

$ cd my-existing-app
$ aes init
$ aes sync
  • Auto-detects your project type and suggests a matching template
  • Customize skills, permissions, and memory for your workflow
  • Merges cleanly with existing CLAUDE.md or .cursorrules
Learn how it works →
Why it matters

Built for how you actually work

For individual developers

Your agent becomes a specialist

  • Templates give your agent domain expertise out of the box — ML, web, DevOps, research
  • Install proven skills from the registry. One command, zero learning curve
  • Memory persists across sessions. Every conversation makes your agent smarter
  • Works across Claude, Cursor, Copilot, Windsurf, Codex, OpenClaw — switch without starting over

For teams

Every AI works from the same architecture

  • Every agent works from the same architecture, conventions, and design principles
  • New teammate? aes init --from your-team/template — instant project context
  • Instructions, permissions, and skills live in .agent/, always in sync via git
  • Ship independently. Everything fits because agents share the same foundation
5 Domain templates
50+ Community skills
6 AI tools supported
1 Standard to learn
The ecosystem

A growing library of agent skills

Install proven skills built by experienced engineers. One command, zero learning curve.

deployCI/CD & infrastructure
code-reviewQuality & best practices
security-auditVulnerability scanning
db-migrateSchema migrations
ml-experimentTraining & evaluation
docs-genAPI documentation
Browse the registry →
See the difference
Without AES
You Deploy the app to staging
Agent What’s your deployment process?
You Docker + GitHub Actions, like I said last time...
Agent What registry do you push to?
You ECR. I already told you this yesterday
Agent What are the environment variables?
...20 messages later, still configuring
With AES
You Deploy the app to staging
Agent Running deploy skill...
Built Docker image app:v2.4.1
Pushed to ECR staging registry
Deployed via GitHub Actions
Staging is live at staging.app.com

Stop re-explaining your codebase to AI

pipx install aes-cli