Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Install and configure Persistia to give Claude Code self-updating persistent memory across sessions
triggers
["set up persistent memory for this project","install persistia brain system","give claude persistent memory","configure self-updating project memory","set up autonomous task scheduling","create persistent brain repository","install persistia for claude code","enable automatic context updates"]
Persistia gives Claude Code self-updating persistent memory by creating a _brain/ directory that tracks your project context, skills, tasks, and operations. It uses git diffs to update only what changed, making it token-efficient and always current.
What It Does
Initial project scan — reads code, docs, configs, environment files on first run
Optional: GitHub CLI (gh) for automatic private repo creation and cloud backup
The setup script:
Creates _brain/ directory structure
Performs initial project scan
Asks for missing context (credentials, product details)
Initializes isolated git repository
Installs scheduler (launchd on macOS, systemd on Linux)
Optionally creates private GitHub remote
Directory Structure
_brain/
├── .git/ # Isolated brain repository
├── index.md # Agent reads this first every session
├── dashboard.html # Live command center (auto-refreshes every 5 min)
├── core/ # Product, brand, ICP definitions
│ ├── product.md
│ ├── brand.md
│ └── icp.md
├── operations/ # Auto-updated metrics and customer data
│ ├── metrics.json
│ └── customers/
├── skills/ # Permanent rules and behaviors
│ ├── code-style.md
│ ├── deployment.md
│ └── ...
└── tasks/ # Scheduled task queue
├── queue.json
└── completed/
Key Concepts
Skills (Permanent Memory)
Skills are persistent rules that apply to all future sessions. Create them conversationally:
Example interaction:
You: "Always use TypeScript strict mode and include JSDoc for public APIs"
Claude: [creates _brain/skills/typescript-conventions.md]
Skills are automatically loaded at session start via index.md.
Tasks (Autonomous Scheduling)
Schedule tasks in _brain/tasks/queue.json:
{"tasks":[{"id":"weekly-metrics","description":"Pull last week's numbers from Stripe, flag anomalies, queue follow-ups for accounts below quota","schedule":"weekly","day":"monday","time":"09:00","batch_size":50,"timeout_minutes":10,"enabled":true},{"id":"inactive-signups","description":"Find signups from last 30 days with no activity. Visit company websites and draft personalized re-engagement emails","schedule":"daily","time":"14:00","batch_size":20,"timeout_minutes":5,"enabled":true}]}
Task properties:
batch_size — cap items processed per run (prevents timeouts)
timeout_minutes — default 5, increase for complex operations
schedule — daily, weekly, monthly
enabled — toggle without deleting
Operations (Auto-Updated Context)
_brain/operations/ contains auto-updated data from your tools:
Persistia reads credentials from .env to update this automatically.
Common Usage Patterns
Initial Setup Conversation
You: "Install persistia"
Claude: [runs setup.sh]
Claude: "I need a few details. What's your product positioning in one sentence?"
You: "Developer platform for real-time data sync"
Claude: "What integrations should I track?"
You: "Stripe for revenue, Intercom for support, PostgreSQL for usage"
Claude: [creates _brain/core/product.md, configures integrations]
Teaching a Skill
You: "When deploying, always run tests, build, then push to staging before production"
Claude: [creates _brain/skills/deployment-process.md]
# Next session
You: "Deploy the latest changes"
Claude: [automatically follows deployment-process.md]
Scheduling a Task
You: "Every Monday at 9am, check which customers had usage drops >30% and draft personalized check-in emails"
Claude: [adds task to _brain/tasks/queue.json with batch_size and timeout]
Asking Cross-Tool Questions
You: "Which customers churned this month and what were their last support interactions?"
Claude: [queries _brain/operations/metrics.json and customer records, cross-references support history]
Viewing Dashboard
Open _brain/dashboard.html in browser — auto-refreshes every 5 minutes showing:
Recent metrics trends
Task queue status
At-risk customers
Pending actions
Configuration
Environment Variables
Persistia reads existing .env files for integrations. Common variables:
# Stripe
STRIPE_API_KEY=sk_live_xxxxx
# Database
DATABASE_URL=postgresql://user:pass@host:5432/db
# Support
INTERCOM_ACCESS_TOKEN=xxxxx
# GitHub (optional, for brain repo backup)
GITHUB_TOKEN=ghp_xxxxx
