AIris Keeper

API key management for the age of AI agents.

AIris Keeper manages API keys for LLM-powered development workflows. Instead of hiding secrets, it minimizes the blast radius when keys leak — because in the age of AI agents, they will.

Core Principles

• Damage control over secrecy — Budget caps, short-lived keys, and automatic rotation mean a leaked key is worth nothing
• Key granularity — Automatically isolate keys by project, environment, and scope. One leak affects nothing else
• Agent-native — Built for Claude Code, Codex, Cursor, and other AI coding agents to use directly via CLI
• Doppler-simple, not Infisical-complex — Minimal surface area, maximum utility

Features

• Vault — AES-256-GCM encrypted secret storage with environment scoping (dev/stg/prd)
• Provisioners — Automatic API key creation via provider Admin APIs (OpenAI, more coming)
• Budget — Per-customer monthly spending limits with alerts
• Rotation — Atomic key rotation: revoke old, create new, update environment in one command
• Injection — Run commands with secrets injected as environment variables
• Audit — Full audit log of all key operations

Quick Start

npx airis-keeper --help

Setup

# Initialize from your .env.example
keeper init --from .env.example

# Connect a provider
keeper provider add openai

# Provision a key with budget cap and TTL
keeper provision --project myapp --env staging \
  --provider openai --scope "chat:gpt-4o" \
  --budget 10 --ttl 24h

Daily Usage

# Inject secrets and run your app
keeper run --project myapp --env staging -- node server.js

# Check status across all projects
keeper status

# Rotate keys atomically
keeper rotate --project myapp --env prod

Secret Management

# Store encrypted secrets
keeper secret set API_KEY sk-xxx --env prod
keeper secret get API_KEY --env prod
keeper secret list --env prod

# Delete a secret
keeper secret delete OLD_KEY --env staging

Budget Management

# Set a monthly budget cap
keeper budget set --customer acme --limit 10

# Check current usage
keeper budget get --customer acme

How It Works

.env.example → keeper init → keeper provision → keeper inject
                                  ↓
                         Provider Admin API
                         (OpenAI, Anthropic, ...)
                                  ↓
                         Key with $10 cap + 24h TTL
                                  ↓
                         Encrypted in vault (AES-256-GCM)

Each key is:

• Budget-capped — $10/month max? Even if leaked, damage stops at $10
• Time-limited — 24h TTL? Key is dead tomorrow regardless
• Scope-restricted — Only chat:gpt-4o? Can't touch embeddings or other models
• Isolated — One project, one environment. Leak one, the rest are safe

Architecture

keeper CLI
├── vault/        — AES-256-GCM encryption, tenant key derivation
├── provisioners/ — Provider adapters (OpenAI, more planned)
├── budget/       — Per-customer spending limits + usage tracking
├── commands/     — CLI command handlers
└── index.ts      — Library exports for programmatic use

Environment Variables

Variable	Required	Description
VAULT_ROOT_KEY	Yes	256-bit root encryption key
SUPABASE_URL	Yes	Supabase project URL
SUPABASE_SERVICE_ROLE_KEY	Yes	Supabase service role key
OPENAI_ADMIN_API_KEY	No	OpenAI Admin API key (for provisioning)
OPENAI_PROJECT_ID	No	OpenAI project ID
KEEPER_ORGANIZATION_ID	No	Default organization ID

License

Elastic License 2.0 (ELv2) — Free for individual and internal use. Cannot be offered as a competing managed service.