NeusymNeuro Symbolic
A2 Agent Framework

Build agents from reusable foundations. Ship them as systems.

A2 is the platform layer for creating AI agents with roles, tools, memory, workflow logic, SDK support, and deployment-ready structure.

01 framework02 SDK03 ecosystem
Data connectorsModel adaptersMemory & stateTools & APIs
A2Framework / SDK

Bring in structured records, documents, events, and private system context.

Single agentsMulti-agent systemsWorkflowsDeployment
Choose a build path

Start with an agent, a team of agents, or a workflow.

Role-first

Single agent

Define one reliable agent with a narrow role, approved tools, memory rules, and deployment configuration.

1 role4 core parts1 deployable unit

A2 Framework & SDK Overview

The development foundation for building AI agents, with the structure, tools, and patterns needed to create single-agent and multi-agent systems.

Why A2 Exists

Building AI agents from scratch is complex. A2 reduces this complexity with reusable building blocks for workflows, memory, tools, models, data access, and deployment.

Agent Building Blocks

A2 gives each agent the core components needed to perform a role: instructions, models, tools, memory, workflows, data access, metadata, and task definitions.

Agent building blocks

1. Instructions & Role

Each agent is defined by a purpose, operating rules, and behaviors expected for a specific task.

2. Models, Tools & Memory

Agents connect to models, call tools, and retain useful context or state to solve problems effectively.

3. Workflows & Data Access

Agents follow steps, access databases or APIs, and participate in broader execution flows.

4. Metadata & Tasks

Capabilities, configuration, and task definitions make agents reusable, searchable, and easier to orchestrate.

Workflow and multi-agent logic

Coordinate agents across a real execution path.

A2 supports ordered stages, delegation, message exchange, and repeatable handoffs from input data to final output.

Step-Based Execution

Agents follow ordered stages, branching logic, and decision points to complete complex work.

Delegation & Collaboration

One agent can assign subtasks to another, pass context forward, and coordinate specialized skills.

Example Flow

A data agent gathers information, an analysis agent interprets it, and a report agent prepares the output.

Input Data
Collector Agent
Analysis Agent
Report Agent
Final Output
Stage 1Input Data

Connect source systems, files, events, and user context before a run begins.

SDK developer experience

A practical layer for building, testing, and registering agents.

The SDK gives teams interfaces, examples, deployment tooling, and packaging patterns so A2-built agents can connect into the wider A3 marketplace.

Libraries & APIs

Developers get ready-to-use interfaces for building agents, connecting models, and invoking tools.

Examples & Templates

Starter patterns and sample implementations make it easier to prototype and learn by doing.

Testing & Deployment

The SDK helps teams move from local development to deployment with confidence and repeatability.

Connect to A3

A2-built agents can be described, packaged, registered, discovered, tasked, and monetized through the wider ecosystem.