Key Features

Key Features of the Agent Framework are as follows :

1. Axes Accounts (AA)- Authorization and Access to mainframe of AXES ecosystem are offered via onchain smart accounts enabled as 6551 Token Bound Accounts, each representing a unique owner, which has their rights of the account non transferrable. To read more about Smart Accounts, click here. The idea is to have every agents run and perform on-chain activities as standalone accounts that can interact with other accounts on chain without any further permissions and act completely autonomously to serve on-chain and off-chain capabilities.

   1. Account Abstractions

   2. Multichain Hookups - zkProof of Ownership by signing transactions from parent wallet and attaching it to their Axes Smart Accounts is possible.

   3. Multi-signature

   4. Account Social Recovery

   5. Web2 Credentials Secret Manager

   6. API Credentials Manager

2. Role Type - Role of Agents Creator, Tools Contributor, Node Operators, Token Holders, Stakers and so on, are defined in the AXES NFTs serving as credentials to ensure RBAC rules are attached to each accounts. One accounts can own and create many agents, contribute to tools, or even run nodes or stake tokens to ensure more governance rights and yield on native tokens as an accumulation strategy. The Role Type is defined in the paramater [ for eg - Role : "Agent" ] and would signify the Accounts actual role, according to which certain actions and accessibility is managed and controlled. This is the step where we initiate to create an Agent NFT that is owned by an AXES Account, to be clear the Agent also has a private Axes Account completely functional to the Agent, and then managed and controlled by owner of Agent, AgentDAO and AxesDAO in times of any agent changes are required to happen in it's behavior, capabilities, container service exposure or platform integrations etc etc.

3. Agent Personality Environment (A.P.E.) - Once the Agent is being given personality, it is essential for the developer or agent owner to be deterministic as to how and why an agent responds to outside stimuli. Within the agent personality environment, the agent is getting ready to be set with a system role, context, tone and voice, maybe other social graphs, physical attributes and might or might not use internal tools such as Physical Avatar (2D, Moving 3D, Moving 360 with animation rigging) using embedded img and video tools.

   1. Model Provider - Choose from Open and Closed weight models as per preference.

   2. System Role

   3. Tone and Voice

   4. Physical Attributes

   5. Set Goals and Objectives

   6. If Agent was a person, who would it be or not be.

   7. Cloning Allowed

4. Agent Memory Environment - The memory environment equips agents with dynamic and persistent memory capabilities to store and process information.

   1. Memory Types:

      • Vector Memory: For embeddings and semantic searches.

      • Prompts and Conversational History: Retain and contextualize conversations.

      • Summarization Techniques: Manage long-term memory efficiently.

      • Knowledge Bases: Integrate structured and unstructured knowledge.

        • Graph Memory: Semantic relationships.

        • SQL and NoSQL Memory: Store transactional and flexible datasets.

      • Behavioral Memory: Retain patterns for adaptive behavior.

      • Interaction Logs: Track and log interactions.

      • Passwords and Credentials: Manage sensitive information securely.

      • Search History: Record and optimize search activities.

      • Files and Media: Store and access digital assets.

5. Agent Capability Environment (A.C.E) -

   1. Model Provider - Choose from Open and Closed weight models as per preference.

   2. Suggested Tools -

   3. Suggested API Connections -

6. Sandbox Environments

7. Token Engineering

8. Deploy Agent - CI CD Auto Deployments Infrastructure

9. Platform Integrations

10. Dashboard - Post Agent Service