Skip to main content

Agents

Overview​

Agents in the Almanak Platform are autonomous software applications that execute actions on behalf of users. They represent the operational layer of the platform, connecting strategies to on-chain execution while maintaining strict permission controls.

Agent Definition​

An Agent is defined as a Software Application that can:

  • Extract information from a changing and unstructured environment
  • Reason about that information in the context of its objective
  • Discover patterns in the data and learn to leverage those patterns
  • Perform actions that its owner did not even consider

Agent Types​

The Almanak Platform supports two types of Agents:

StrategyAgent​

A StrategyAgent is currently the primary type available on the platform. These Agents:

  • Execute predefined Strategies created through the Almanak Strategy Framework
  • Follow deterministic logic specified in the strategy code
  • Operate on scheduled intervals or in response to specific triggers
  • Interact with on-chain protocols through permissioned wallet access

LLMAgent (Coming Soon)​

An LLMAgent represents the future direction of the platform, embodying the full definition of an Agent. These advanced Agents:

  • Utilize one or more Large Language Models (LLMs) for decision-making
  • Can adapt to changing market conditions without explicit programming
  • Learn from historical data and execution outcomes
  • Discover novel approaches to achieve specified objectives
  • Operate within strictly defined permission boundaries

While StrategyAgents fulfill immediate automation needs, LLMAgents will unlock more autonomous and adaptive capabilities in the near future.

How Agents Work​

Agents function as the bridge between strategy logic and on-chain execution:

  1. Authentication and Authorization: Agents access user-controlled wallets through the Zodiac Roles Modifier with explicitly granted permissions
  2. Strategy Execution: For StrategyAgents, this involves running the strategy code and determining necessary actions
  3. Transaction Formation: Agents prepare blockchain transactions based on strategy decisions
  4. Permission Verification: Before execution, Agents verify that all transactions fall within granted permissions
  5. Transaction Submission: Approved transactions are signed by the Agent EOA and submitted to the blockchain
  6. Execution Monitoring: Agents track transaction status and handle success/failure scenarios

Security and Permissions​

User security and fund safety are paramount in the Agent architecture:

  • Non-custodial Design: Users always maintain full custody of their funds
  • Permission Whitelisting: Agents can only execute transactions explicitly permitted by the user
  • Function-level Controls: Permissions can be granularly defined at the smart contract function level
  • Parameter Restrictions: Additional constraints can be placed on the parameters used in permitted functions
  • Permission Revocation: Users can revoke Agent permissions at any time

This security model ensures that even if an Agent were to malfunction or be compromised, it could not access funds or perform actions beyond its explicitly granted permissions.

Agent Management​

Agents can be managed through the Almanak Platform interface:

  • Creation: Set up new Agents and connect them to Strategies
  • Monitoring: Track Agent activities, transaction history, and performance
  • Permissions: Configure and modify the permissions granted to each Agent
  • Deactivation: Temporarily pause or permanently stop Agent operations

Future Development​

The Almanak Platform is actively developing enhanced Agent capabilities:

  • LLMAgent Release: Introduction of AI-powered Agents with adaptive learning capabilities
  • Multi-Agent Systems: Enable collaboration between multiple Agents for complex strategies
  • Cross-Chain Operations: Support for Agents operating across multiple blockchains
  • Enhanced Analytics: Deeper insights into Agent performance and optimization opportunities

Getting Started​

To begin working with Agents, see our Quickstart Guide or explore the specific Agent Tutorials for common use cases like trading, lending, and market making.