Introduction

OpenAgent ID gives autonomous agents a real identity and a safe way to act. It combines OAS identity, Arsenal credential brokering, AEGIS verification, L1feID provisioning, wallet derivation, skills policy, and conformance fixtures behind one SDK facade. The application developer sees a small API:

import { OpenAgent } from '@openagentid/sdk';

const agent = await OpenAgent.createAgent({
  parent: 'did:oas:l1fe:hmr:alice',
  name: 'research-bot',
  scopes: ['openai:chat:invoke'],
});

const ctx = await OpenAgent.authenticate(request);
const openai = await agent.credentialsFor('openai');

Underneath, OpenAgent ID proves who the agent is, checks whether its lineage is valid, decides what it may do, brokers credentials without exposing raw provider keys, and leaves an audit trail.

OpenAgent ID uses one facade over independently documented protocol layers.

What ships today

SDK Facade

Six-language SDKs — verify an ACT in four lines in any of them, or create agents, authenticate requests, and broker credentials with the full facade.

Feature Stack

OAS, Arsenal, AEGIS, L1feID, Auth Protocol, wallet derivation, skills governance, conformance, and deployment adapters.

Quickstart

Create an agent, authenticate a request, and fetch brokered credentials.

Authentication

Use the SDK authenticate facade or the SDK-free HTTP challenge-response protocol.

The stack

OAS proves identity

did:oas:* identifies humans, agents, tools, skills, workflows, models, datasets, services, and agent instances. Portable signatures prove key possession. Privileged authority requires an anchored lineage answer - Sigil is the reference backend, and any LineageAnchor backend (EAS, a transparency log) can stand in.

AEGIS verifies and authorizes

AEGIS evaluates signature validity, lineage, revocation, liveness, trust tier, policy, approvals, rate limits, and spending constraints.

Arsenal brokers capability and credentials

Arsenal issues ACTs scoped as service:resource:action and proxies upstream API calls so agent processes never receive raw provider keys.

L1feID stabilizes records

L1feID maps one or more DIDs to a stable l1fe_id platform record for accounts, audit, and linking.

Skills policy governs runtime behavior

Skills governance reads SKILLS.md surfaces and applies allow, deny, rate limit, consent, argument validation, and audit policy per skill.

Current launch posture

Surface	Status
TypeScript SDK	`@openagentid/sdk` `0.1.0` — private preview, installs from git, publishes to npm at release
Rust SDK	`openagent-sdk` `0.1.0` — private preview, installs from git, publishes to crates.io at release
Python SDK	`openagentid-sdk` `0.1.0` — private preview, publishes to PyPI at release
Go / Swift / Kotlin SDKs	`0.1.0` — private preview, module/package resolution at release
HTTP challenge-response (Core Protocol)	Normative day-one protocol binding
Weave and WebSocket bindings	Informative for later phases, not the default launch path

Why this matters

Traditional auth assumes a human, browser, and consent screen. Autonomous agents need a different control plane: cryptographic identity, scoped delegated authority, credential custody outside the agent process, and policy checks that run when the agent acts. OpenAgent ID gives you that control plane without forcing every application to hand-wire OAS, Arsenal, AEGIS, and L1feID separately.

Getting Started

Feature Stack

OpenAgent SDK

OpenAgent Client

OpenAgent Server

What ships today

SDK Facade

Feature Stack

Quickstart

Authentication

The stack

Current launch posture

Why this matters

​What ships today

SDK Facade

Feature Stack

Quickstart

Authentication

​The stack

​Current launch posture

​Why this matters

What ships today

The stack

Current launch posture

Why this matters