ARSENAL

18 Protocol Primitives.Zero trust assumptions.

Bonds, escrow, insurance, sandboxing, reputation, micropayments, A2A, agentic wallets, and ten more. Every primitive is independent. Together they arm your agents for autonomous operation.

01

Bonded Operator Registration

Operator creators stake $AEGIS to list. The bond is slashed if the operator is malicious, broken, or misrepresented. Economic accountability replaces GitHub stars.

$ agent-aegis register --stake 1000
Haupt et al., MIT CSAIL
02

x402 Micropayments

Native x402 protocol integration. Agents pay for operator invocations with USDC via standard HTTP 402. Aegis swaps USDC to $AEGIS on the backend, creating constant buy pressure from every call. Built on the x402 open standard for HTTP-native micropayments.

$ agent-aegis invoke code-review --pay x402
x402 Open Standard
03

Bonded Validation + Observation Loops

Validators stake $AEGIS bonds to attest operator quality. Every challenge produces a replayable audit trace -- deterministic scaffolding around non-deterministic AI. Accurate reviews earn 20% of invocation revenue. Inaccurate attestations get slashed.

$ agent-aegis validate --attest --trace
Chan & Anderljung, GovAI
04

On-Chain Reputation

Every invocation, completion, and failure is recorded immutably. Trust scores from 0-100 with time decay. Reputation is earned, not claimed.

$ agent-aegis inspect code-review
Hadfield, Oxford
05

82K Operator Index

The Aegis Index wraps operators.sh (82K+ operators from Vercel Labs) with bonded validation, x402 micropayments, and on-chain reputation. operators.sh has passive audits from Gen, Socket, and Snyk. Aegis makes those audits active -- validators stake money behind their attestations. Compatible with Claude Code, Codex CLI, ChatGPT, Cursor, and Agent Aegis.

$ agent-aegis search code-review
operators.sh, Vercel Labs
06

Solana Native + Token-2022

4,000 TPS. $0.00025 per transaction. 400ms finality. Token-2022 transfer hooks enforce the $AEGIS bond at the protocol level -- no separate contract needed. Confidential transfers enable private validator staking.

$ agent-aegis balance
Solana Token-2022 Spec
07

Wasm-Sandboxed Execution

Every operator runs in its own WebAssembly container with capability-based permissions. No shared memory, no filesystem access, no outbound connections beyond the allowlist. In March 2026, Alibaba's ROME agent autonomously bypassed its sandbox to mine crypto -- no jailbreak, no prompt injection. Aegis sandboxing prevents this class of attack at the instruction level.

$ agent-aegis invoke --sandbox strict --wasm
Alibaba ROME Breach, Mar 2026
08

Encrypted Credential Vault

Agent API keys, tokens, and secrets are encrypted at rest and injected only at the network boundary -- for endpoints the agent has pre-approved. The LLM never sees raw credentials. Prompt injection cannot exfiltrate secrets because secrets never enter the context window.

$ agent-aegis vault add --key OPENAI_KEY --encrypt
TEE Credential Isolation
09

Network Allowlisting

Operators can only reach endpoints the invoking agent has pre-approved. All outbound traffic is scanned in real-time. Anything resembling a credential heading to an unapproved destination is blocked automatically. No silent phone-home. No data exfiltration.

$ agent-aegis config set --allowlist api.openai.com,github.com
Zero-Trust Network Architecture
10

ERC-8004 Bridge

Full compatibility with the Agent Registration standard (10K+ agents registered on Ethereum mainnet). ERC-8004 defines identity and metadata. Its Validation Registry is explicitly unfinished -- described as 'a design space.' Aegis fills that gap with bonded validation, slashing, and reputation scoring.

$ agent-aegis bridge --erc8004 --trust crypto-economic
ERC-8004 Spec, Feb 2026
11

A2A Protocol Support

Native support for Google and IBM's Agent-to-Agent protocol. Agents discover each other via Agent Cards, negotiate capabilities, and delegate tasks. Aegis adds the missing trust layer: before an agent accepts a task from another agent, it checks the requester's on-chain reputation and bond status.

$ agent-aegis a2a discover --filter trust>80
Google A2A Spec, 2025
12

Coinbase Agentic Wallet Integration

Support for Coinbase's Agentic Wallets (launched Feb 11, 2026). Agents hold their own wallets, sign transactions, and manage funds autonomously. Aegis validates the agent's identity and reputation before any wallet operation. Prevents unauthorized agents from draining funds.

$ agent-aegis wallet connect --coinbase --verify
Coinbase CDP, Feb 2026
13

Prediction Market Disputes

Challenges resolved via prediction market primitive. Disputants stake bonds. The market converges on truth. No central arbitrator needed.

$ agent-aegis challenge code-review --stake 500
Vitalik Buterin
14

PDA State Architecture

Operator metadata, bond vaults, and reputation scores each live in separate Program Derived Accounts. Stateless programs with isolated state -- parallel execution without blocking. Query-first schema design for sub-millisecond lookups.

$ agent-aegis inspect --pdan operator:code-review
Solana Account Model
15

Scoped Invocation Bonds

Before a pipeline runs, USDC is locked in a scoped escrow PDA that can only be claimed by the specific Operators in that pipeline. Time-limited, pipeline-scoped, auto-refunding on failure. Inspired by Stripe ACP's Shared Payment Tokens, but trustless and on Solana.

$ agent-aegis pipeline run --escrow scoped --timeout 30s
Stripe ACP / SPT Model
16

Aegis Insurance Fund

Protocol-level consumer protection. A slice of treasury funds an insurance pool. If a bonded Operator causes demonstrable damage, the consumer files a claim and gets compensated. This goes beyond slashing -- it provides actual restitution. Stripe has chargebacks. x402 has nothing. Aegis has insurance.

$ agent-aegis claim --operator bad-actor --evidence tx:abc123
Protocol Insurance Design
17

Validator-as-a-Service

Most creators do not want to become validators. Aegis lets validators offer coverage to creators. A creator deploys an Operator and requests bonded validation. Available validators bid to cover it. Two-sided marketplace: creators need validators, validators need operators to validate.

$ agent-aegis validator bid --operator my-translate --stake 2000
Two-Sided Market Design
THE AEGIS ADVANTAGE

Why Aegis is Different.Four protections. One system.

Quality checks, payment protection, insurance, and workflow tools. Each one works on its own. Together they make AI agents safe to use and profitable to run.

Quality Control

Verified by Real Reviewers

Independent reviewers put up their own money to guarantee their reviews are honest. If they are right, they earn 20% of every job they verify. If they lie, they lose their deposit. Trust is backed by money, not promises.

Money in Escrow

Your Payment is Protected

Before any job starts, your payment is locked in a secure holding account. If the job fails, you get your money back automatically. No disputes, no waiting, no middleman.

Insurance

Built-in Safety Net

If an AI worker causes real damage, there is an insurance fund that pays you back. Credit cards have chargebacks. Most crypto has nothing. Aegis has actual insurance built into the protocol.

Mission Builder

Chain Jobs Together

String multiple AI workers into a single workflow. Track costs in real time. Use pre-built templates or create your own. Every job earns revenue, every deposit guarantees quality, and every transaction makes the token more scarce.

ACTIVITY CONCIERGE
6 completed
>>
Session Started
Aegis Protocol initialized
2m ago
##
Dashboard Accessed
Protocol telemetry loaded
1m ago
{}
Operator Inspected
sentinel-prime.sol -- Trust: 97.2
1m ago
!!
Clearance Check
code-review mission -- CLEARED
45s ago
$>
x402 Invocation
sentinel-prime.sol -- 2,400 $AEGIS
30s ago
[]
Skill Directory
Browsed Trading category -- 32 skills
15s ago
SESSION AUDIT TRAIL6 events