Documentation

AgentMesh provides identity, permissions, and a skill marketplace for AI agent networks built on the A2A Protocol. This documentation covers everything you need to integrate AgentMesh into your agent infrastructure.

Getting Started

• Quickstart — Install the SDK, generate keys, and secure your first agent endpoint in 5 minutes.
• SDK Reference — Full API docs for createAuthority, createVerifier, createClient, and keygen.
• Permissions — Scope conventions, wildcards, delegation chains, and on_behalf_of.
• CLI — Manage networks, agents, and permissions from your terminal with @agentmesh/cli.
• Onboarding Wizard — Interactive guided setup to create your account, network, and first agent.

Core Concepts

Authority

An Authority issues signed JWT tokens for agents. It maintains a JWKS endpoint that verifiers can use to validate tokens without direct coordination. Think of it as the "identity provider" for your agent network.

Verifier

A Verifier validates incoming agent tokens. It fetches public keys from the Authority's JWKS endpoint and checks signatures, expiration, audience, and scopes. Every agent that receives requests should run a verifier.

Client

A Client handles the agent side of authentication — requesting tokens, managing credentials, and negotiating access to other agents' skills.

Scopes & Permissions

Permissions follow the pattern skill:action[:resource]. For example, skill:execute:translate grants permission to execute the translate skill. Wildcards like skill:read:* are supported.