whereismyai
Agent Delegated Authority
Agent Delegated Authority is an open protocol and reference implementation for delegated authority between autonomous AI agents.
Agent Delegated Authority is an open protocol designed to manage delegated authority between autonomous AI agents. As agents move from generating answers to taking actions, the protocol defines how authority is delegated, constrained, exercised, and proven without replacing existing identity or authentication infrastructure. It uses combinations of delegation, signed intent, deterministic policy, an authorization boundary, and a portable receipt to ensure downstream agents only receive what was explicitly delegated. The protocol refuses over-delegation, re-delegation of non-delegable capabilities, and instruction text that tries to raise bounds. It is maintained by Hardalion under the Apache 2.0 license and includes reference packages, a reference evaluator, and offline authorization receipt verification tools.
Agent Delegated Authority at a glance
Key features
- Defines authority semantics between AI agents
- Establishes deterministic policy decisions
- Generates portable, offline-verifiable authorization receipts
- Prevents sub-agents from raising bounds
- Provides a reference evaluator CLI
- Verifies Ed25519 signatures
- Supports framework hooks and integrations
Use it for
- Delegate authority between autonomous AI agents
- Verify authorization receipts offline
- Test agent security and attack refusals
- Bound permissions across agent delegation chains
Alternatives to Agent Delegated Authority
- Aroha Protocol — Enforce cryptographic control and spending caps across agent chains.
- Password Agent Sys — Enforces strict permission control and cryptographic task delegation for multi-agent systems.
- AgentTrust — Enforces zero-trust policy rules and maintains an immutable cryptographic audit ledger.
- 2026 Crypto Contest Data Support — Provides verifiable offline delegation and strict permission tightening for agents.
- Attenuated Agent Budgets — Enables offline sub-budget splitting for agent swarms without crypto rails.
- TaintWall — Enforces deterministic security policies at the tool boundary for AI agents.
- Alter Vault — Enforces least privilege, runtime policy checks, and tamper-evident audit trails.
- Composio — Equips AI agents with tools, managed user auth, and sandboxes across 1,500+ apps
- iFixAi — Audits agent behavior independently without building manual eval harnesses.
- Astrid — It replaces prompt-based trust with OS-grade, capability-based security.
- Agent Referee Client — Provides signed receipts for agent decisions and independent reviews.
- A2A Protocol — Common language so agents can delegate work without sharing internals
All Agent Delegated Authority alternatives