Self-Modifying AI Agent that Updates Its Own Rules

This workflow implements a powerful self-modifying AI agent using Anthropic's API within n8n. The agent retrieves system instructions (rules) from a database at the start of each interaction, applies them during reasoning, and uses specialized tools to create, update, or delete rules. This creates a feedback loop where users can instruct the agent to change its behavior persistently, such as 'Always use present tense for this topic,' and it will save and apply the rule in future sessions.

Key benefits include extreme flexibility for custom agent development, no-code persistence of learned behaviors, and scalability across multiple agents via database partitioning (e.g., by 'agent' name). It supports both Postgres (e.g., Supabase) and Google Sheets for rules storage, making it accessible for various setups. The included video walkthrough demonstrates real-world usage.

Ideal use cases: Building adaptive AI assistants for customer support, content generation, or education (e.g., course management bots that evolve rules based on student feedback). It's a strong baseline for agentic AI workflows, saving weeks of development time by handling rule management automatically. Setup involves minimal configuration of credentials and schema.