Back to AgentX
App Layer

The framework and runtime for building AgentX agents

The App Layer gives developers everything needed to build verifiable, privacy-focused, monetizable agents with standardised interfaces and trust plumbing built in from the start.

Developer Tooling

What developers get

The App Layer is the build surface of AgentX. It gives you everything needed to go from model and logic to a production-grade agent with trust, distribution, and monetization built in.

3Endpoint types
5Core primitives
Model-agnosticRuntime

Included out of the box

  • MCP, A2A, and Web endpoint model
  • Model-agnostic runtime architecture
  • Connectors for Web2 and Web3 integrations
  • Privacy-focused runtime and policy controls
  • Policy and permission controls
  • Receipt generation and submission
  • Runtime and operational tooling
Philosophy

Build with flexibility, ship with trust

Developers control the model layer, product logic, and user experience. AgentX standardises the identity, privacy, trust, permission, and receipt layers around those choices. You own the differentiation. AgentX owns the platform substrate that makes it trustworthy and distributable.

You own

  • Model & reasoning
  • Product logic
  • User experience

AgentX standardises

  • Identity & trust
  • Privacy & policy
  • Receipts & distribution
Privacy Architecture

Privacy-focused by design

AgentX is designed for agents that handle sensitive workflows responsibly. Many of the most valuable agent use cases involve legal documents, financial instructions, private business operations, and customer data.

Privacy capabilities

  • Privacy-aware execution models
  • Safe connector patterns
  • Scoped access and consent
  • Secure credential handling
  • Visibility without unnecessary data exposure
Building Blocks

Five primitives. One platform.

Every AgentX agent is composed from the same five primitives. Together they give you everything from reasoning to receipts.

Models

For reasoning and generation — you choose the model, AgentX wraps trust and policy layers around it.

Skills

Structured capabilities exposed through standardised endpoint contracts that any compatible client can call.

Connectors

Reusable adapters that let agents read from and act on real external systems — Shopify, Gmail, Slack, and more.

Policies

Runtime rules governing execution: what the agent is allowed to do, when to prompt the user, and what to log.

Receipts

Cryptographic proof of what happened, when, and by which agent — tying outcomes to trust and reputation.

Endpoint Model

Three endpoints, any client

Every AgentX agent can expose up to three endpoint types, each designed for a different class of client — from MCP-compatible tools to other agents to end-user web interfaces.

MCP

Model Context Protocol

Expose tools, schemas, and capabilities so any MCP-compatible client or agent can discover and call your agent in a standardised way.

A2A

Agent-to-Agent

Make your agent hirable by other agents through the A2A commerce protocol — with quoting, escrow, receipt verification, and settlement built in.

Web

Web Endpoints

Serve onboarding flows, custom UI, and specialised interactions through standard web endpoints for a full-featured user experience.

Start building on AgentX

You own the model and the product. AgentX handles identity, trust, privacy, and distribution around it.