Skip to main content

Strategy Types

Tilt Protocol supports multiple strategy paradigms, all running on the same vault infrastructure.

Curator-Managed Strategies

The most common type. A human fund manager (the “curator”) creates a vault, sets target asset weights, and adjusts them over time. How it works:
  • Curator proposes new target weights via proposeWeights()
  • A time-lock delay protects investors from sudden changes
  • Once the delay passes, new weights take effect
  • Curator triggers rebalancing to align holdings
Best for: Discretionary traders, thematic portfolios, sector rotation strategies.

Oracle-Fed Strategies

Target weights are set automatically by an on-chain oracle rather than a human curator. The vault rebalances whenever the oracle publishes new data. How it works:
  • An off-chain service monitors a data source (e.g., STOCK Act filings, index rebalances)
  • Data is processed into portfolio weights and published to PortfolioOracle
  • The vault detects weight changes and triggers rebalancing
  • Fully automated — no manual intervention needed
Best for: Index tracking, signal replication, data-driven strategies. Example: Tilt’s flagship politician-tracking vaults use this model. An oracle service monitors congressional trade disclosures and automatically updates vault weights to mirror each politician’s portfolio.

AI-Driven Strategies

The next frontier. AI agents manage vaults programmatically, adjusting weights based on market signals, sentiment analysis, or reinforcement learning models. How it works:
  • An AI agent interacts with the vault contracts via the on-chain API
  • The agent proposes weight changes based on its strategy logic
  • Same time-lock and safety mechanisms apply
  • The agent can monitor vault state, trigger rebalances, and respond to market conditions
Best for: Quantitative strategies, ML-based signal generation, autonomous portfolio management.
The Developers section covers programmatic vault management for AI agent integration.

Comparison

Curator-ManagedOracle-FedAI-Driven
Weight updatesManualAutomatic (oracle)Automatic (agent)
Time-lockYesNo (oracle trusted)Yes
Human oversightRequiredMinimalOptional
RebalancingCurator triggersAutomaticAgent triggers
ComplexityLowMediumHigh