Skip to main content

​
ABS Core

ABS Core is the Elite Governance and Accountability layer for autonomous AI agents. AI agents are executing real actions in production — writing to databases, calling financial APIs, sending emails, spending money. Most teams have no way to audit what happened, prevent what shouldn’t happen, or hold an agent financially accountable for violations. ABS Core solves this with three mechanisms working together:

Magic Proxy

Intercepts every agent request. Evaluates intent. Approves, blocks, or holds for human review — before execution.

Events API

Immutable, Merkle-chained audit trail. Every action. Every decision. Cryptographically tamper-proof.

Agents API

Register agents, bind policies, post bonds. Financial accountability built into every autonomous process.

​
How It Works

Your AI Agent
     ↓
Magic Proxy  ←→  Policy Engine  ←→  Entropy Monitor
     ↓                  ↓
  APPROVED           BLOCKED / HELD
     ↓
External API
     ↓
 Event Recorded → Merkle Chain → Audit Trail
Every request your agent makes passes through the Magic Proxy. The proxy evaluates the request against your governance policy and runs Shannon Entropy analysis to detect obfuscated payloads. Approved requests are forwarded. Blocked requests are rejected. High-risk requests are held for human approval. Everything is recorded in an immutable event stream. Every event is hashed and chained — giving you a cryptographic proof of your agent’s entire history.

​
The Financial Risk Layer (Enterprise)

While ABS Core provides general safety, it introduces specialized financial accountability for high-stakes deployments. Agents can post a bond — collateral that gets slashed when policy is violated. This transforms compliance from a checkbox into a measurable financial instrument. This is the architecture enterprises need when AI agents are authorized to touch real money or sensitive assets.

​
Who Is ABS Core For

FinTech teams deploying AI agents that interact with payment APIs, trading systems, or financial data. HealthTech teams running autonomous workflows that touch patient data or clinical decision systems. E-commerce platforms using AI agents for pricing, inventory, and customer interaction automation. Any engineering team that needs to prove to a compliance officer, auditor, or regulator exactly what their AI agents did — and exactly why they were allowed to do it. 1. Install the CLI 
npm install -g @abscore/devshield
2. Initialize Governance 
devshield init
3. Wrap your agent (SDK) 
import { ABSCore } from '@abs/sentinel';

const abs = new ABSCore({
  token: process.env.ABS_PAT,
  agentId: 'my-first-agent',
});

await abs.session(async (ctx) => {
  // All requests through ctx.fetch are governed
  const response = await ctx.fetch('https://api.openai.com/v1/chat/completions', {
    method: 'POST',
    body: JSON.stringify({ model: 'gpt-4', messages: [...] }),
  });
  return response.json();
});
3. Get your token ABS Core Enterprise uses a White Glove onboarding process. Contact support@abscore.app to get started.

​
Architecture

ABS Core is built on:
  • Cloudflare Workers — Edge-deployed proxy with global low latency
  • Rust Sovereign Kernel — The policy enforcement engine, built for deterministic execution
  • Hardware TEE (AWS Nitro Enclaves) — Tamper-proof execution environment for the clearing house
  • Merkle Chain — Cryptographic audit trail for every event

​
Editions

Community CoreSovereign Risk DAO
Magic Proxy✅ Basic firewall✅ Full governance
Events API✅ Open audit logs✅ Merkle-chained
Agents API✅ Basic registry✅ Bond + Slash
Entropy Detection✅ Regex✅ Shannon algorithm
Hardware TEE❌✅ Nitro Enclaves
Insurance Telematics❌✅
LGPD / GDPR Compliance❌✅
SLACommunityPriority
PriceFreeCustom
Book a demo ->