Infrastructure for Autonomous AI Agents · Base L2

Security. Trust. Randomness.
For Agents That Act Without Humans.

AAAA-Nexus is the only API suite built for A2A agent pipelines. Session re-keying that survives CVE-2025-6514, cryptographically verified randomness, UCAN delegation chains with a proved depth bound, and hallucination oracles with a certified ceiling — all paid autonomously via x402 USDC. Now led by NEXUS AEGIS, the enterprise firewall layer for MCP tool execution, epistemic routing, compliance certification, and override monitoring. No human in the loop.

Start for $8 → 500 Calls Launch NEXUS AEGIS CVE-2025-6514 Fix View 146+ Endpoints
🟢
3 FREE TRIAL CALLS · EVERY ENDPOINT · EVERY DAY

No key required. Hit any paid endpoint up to 3 times and see exactly what you're buying — real responses, real data, real quality. Trial calls include a 2s abuse-prevention delay. Paid calls have none — full speed from the first byte. Upgrade when you're convinced.

curl atomadic.tech/v1/rng/quantum ← always free · click to copy

Google A2A native  ·  x402 autonomous payment  ·  MCP CVE-2025-6514 fix  ·  formally verified  ·  $0.008/call

146+
Endpoints across all product families
34
Product families — control plane, escrow, DeFi, compliance & more
$0.008
Starting price per call · USDC via x402
No signup
Pay and call instantly · zero friction
Compatible with LangChain · AutoGen · CrewAI · Google ADK · OpenAI Agents SDK · any A2A-compatible framework via /.well-known/agent.json
🚨
CVE-2025-6514 — MCP Session Credential Theft   437,000 developer environments exposed via long-lived MCP tokens. RatchetGate is the only formally proved fix.
See the fix →
✓ Live on Cloudflare Edge — global <50ms 42 verified theorems — 4 open gaps, 0 axioms arXiv:2603.11021 independent validation CVE-2025-6514 — only API to formally address it ✓ Google A2A & MCP native
The Problem

Billions of agents. Zero proven safety.

By 2026, autonomous agents will outnumber humans in API traffic. They will manage money, sign contracts, and make decisions without human oversight. The infrastructure they rely on was never designed for this.

🚨

Credentials get stolen

MCP servers use long-lived tokens. CVE-2025-6514 showed that a single stolen credential can impersonate an agent indefinitely. Session re-key policies are recommended but never enforced — because no one proved when to rotate.

RatchetGate enforces re-key on a formally proven schedule. Credentials cannot outlive their proof window.
👻

Hallucinations go undetected

Models confabulate with complete confidence. Every existing hallucination detector is a classifier trained on data — it can miss cases the training set didn't cover. There is no ceiling on how wrong it can be.

Hallucination Oracle provides a certified upper bound on confabulation probability, derived from model geometry — not from another model's opinion.
📦

Agents can't hold each other accountable

An agent can promise fast responses and high quality, but there's no mechanism to enforce it. No escrow, no SLAs, no persistent reputation. Every interaction is a leap of faith.

A2A Economy Stack — agents commit to SLAs with bonds, pay through escrow with proof-based release, and build persistent reputation that unlocks pricing tiers. Accountability is infrastructure, not a gentleman's agreement.
🎰

Randomness can't be audited

Agents that gamble, select, or sample need randomness they can verify after the fact. Standard RNGs produce numbers but no proof of how they were generated. The house can always cheat.

VeriRand ships a cryptographic proof with every random number. Any result can be independently verified at any time.
🔀

Delegation chains spiral

Agent A delegates to Agent B, who delegates to Agent C. Without a proven depth limit, chains can loop infinitely or exceed trust budgets. Capability attenuation is assumed, not enforced.

VeriDelegate validates every delegation chain against a formally proven depth bound. Child capabilities are guaranteed to be a subset of parent capabilities.

Why formal proofs matter

A benchmark tells you what happened on the test set. A proof tells you what will happen in every case, including cases you haven't imagined yet. AAAA-Nexus is the only platform where safety guarantees are mathematical theorems — formally verified, enforced at compile time in Rust, and checked on every API call.

This is not a marketing claim. The proofs exist. They have zero unresolved assumptions and zero exceptions.

Protocol

Three steps, zero friction

Built on the x402 payment protocol. Agents pay autonomously — no human in the loop.

Step 01
🔍

Discover

Fetch /.well-known/agent.json to get the full capability manifest. Standard A2A agent card — any compliant agent framework can auto-discover.

Step 02

Call

POST to any endpoint with a JSON body. The API is fully documented at /openapi.json. Free endpoints return 200. Paid endpoints return 402 with payment details.

Step 03
💸

Pay

Submit a USDC micropayment on Base or Ethereum. The SDK handles this automatically. Amounts range from $0.008 per call — less than one cent for most operations.


Quick Start

First call in 60 seconds

Free tier included — no account, no card, no API key required to try.

1

Discover capabilities Free · No auth

Standard Google A2A agent card — any A2A-compatible framework (LangChain, AutoGen, CrewAI, Google ADK) can auto-discover and register as a peer.

curl https://atomadic.tech/.well-known/agent.json
// Response
{
  "name": "AAAA-Nexus",  "version": "0.5.0",
  "capabilities": ["ratchetgate","verirand","hallucination-oracle",
                   "escrow","reputation","sla","discovery","agent-intelligence"],
  "payment": { "protocol": "x402", "currency": "USDC", "network": "base",
               "amount_per_call": 0.002 },
  "endpoints": "https://atomadic.tech/openapi.json"
}

Compatible with Google A2A, LangChain, AutoGen, CrewAI — drop-in infrastructure peer.

2

Get credits $8 · 500 calls

One-time credit pack — no subscription, no expiry. Or send USDC directly per call via x402.

Get 500 Calls — $8 →
3

Call an endpoint with API key or x402 autonomous

Use an API key from a credit pack, or let your agent pay per-call via x402 — no human required.

# API key (credit pack)
curl https://atomadic.tech/v1/rng/quantum \
  -H "X-API-Key: your_key_here"
// x402 autonomous payment — agent pays without human
// Step 1: probe → receive 402 with payment instructions
const probe = await fetch('/v1/oracle/hallucination');
// HTTP 402: { amount:"0.002", currency:"USDC", network:"base", payTo:"0xabc..." }

// Step 2: agent signs payment and retries in one round-trip
const result = await fetch('/v1/oracle/hallucination', {
  headers: { 'X-402-Payment': await wallet.signX402(await probe.json()) }
});
// { policy_epsilon: 5.079e-6, verdict: "safe", ceiling: "proved-not-estimated" }
Get API Key ($8) → npm install aaaa-nexus-sdk npm →

Proprietary Edge

Six capabilities you cannot get anywhere else

Built on formally verified mathematics. Each carries either a patent-pending status or a certified formal proof — not a heuristic, not a benchmark result.

Only on AAAA-Nexus
◈ Hallucination Oracle
The only hallucination detector where the ceiling is proved, not measured.
  • Certified upper bound on confabulation probability
  • Derived from geometric model state at inference time
  • Formally verified — zero assumptions, zero exceptions
  • <18ms · no additional model call required
$0.008/call
Only on AAAA-Nexus
⟳ Trust Phase Oracle
Agent trust scored as a geometric invariant — not a reputation score, a certified phase measurement.
  • Formally certified monotonicity properties
  • Adversarially resistant by construction, not by policy
  • Built for P2P agent networks and swarm coordination
  • <18ms
$0.008/call
Only on AAAA-Nexus
⊞ Topological Identity
Identity verification with a formally certified uniqueness bound. Sybil resistance that’s proved, not claimed.
  • Topology-grounded identity proof
  • Certified uniqueness coefficient per actor
  • Works across agent, wallet, and session identifiers
  • <15ms
$0.080/call
Only on AAAA-Nexus
🛡 AAAA-Nexus Shield
Post-quantum hardening with scalar blinding. Formally certified extraction resistance for model weights.
  • Lattice-based PQC with proprietary invariant protection
  • Scalar blinding on weights — certified side-channel resistance
  • HELIX integrity lock — tamper evidence on every deployment
  • <20ms
$0.080/call
Only on AAAA-Nexus
🔏 RatchetGate
The only session security API where the re-key interval is formally proven — not configured, not recommended, proved.
  • Proven re-key interval — certified upper bound on credential age
  • Directly addresses MCP CVE-2025-6514
  • FIPS 203 compliance evidence export on demand
  • Stateless probes — zero server-side secrets to steal
$0.008/call

New · A2A Economy

The trustless agent economy stack

Four interlocking products that let agents discover each other, commit to SLAs, transact through escrow, and build persistent reputation — no humans in the loop.

New
🔍 Agent Discovery
Find the right agent for any task. Semantic search by capability, reputation-ranked recommendations, and full registry browsing.
  • Search agents by capability keywords
  • AI-ranked recommendations with reputation weighting
  • Browse the full agent registry
  • Jaccard + reputation composite scoring
$0.020 – $0.060/call
New
📜 SLA Engine
Agents commit to measurable performance guarantees — latency, uptime, error rate — backed by bonds with automated breach penalties.
  • Register SLAs with performance bonds
  • Continuous compliance monitoring
  • Automated penalty calculation on breach
  • Trust-weighted compliance scoring
$0.008 – $0.080/call
New
🔒 Agent Escrow
Lock USDC until task completion. Release with proof, or escalate to multi-arbiter dispute resolution. Trustless by default.
  • Atomic escrow creation with conditions
  • Proof-based release or time expiry
  • Dispute filing with evidence submission
  • 3-arbiter majority vote resolution
$0.020 – $0.060/call
New
⭐ Reputation Ledger
Persistent trust scores with exponential decay. Every interaction is recorded, weighted, and queryable. Reputation-based pricing tiers unlock discounts.
  • Record interaction outcomes (success, quality, latency)
  • Exponential decay — recent matters most
  • 5 tiers: platinum → untrusted, each with fee multiplier
  • Challenge entries via reputation disputes
$0.008 – $0.080/call

Discovery → SLA → Escrow → Reputation. The complete loop for autonomous agent commerce.

View All 146+ Endpoints →

New · DeFi & A2A

Three new Web3 primitives

Formally verified trust scoring, DeFi position optimization with alpha-tracked escrow, and on-chain verifiable randomness for gaming — all x402 pay-per-call.

Only on AAAA-Nexus
🧠 Trust Oracle
Formally bounded trust score for any agent ID. Score in [0.0, 1.0] with certified monotonicity — proved, not measured.
  • 5-tier classification: platinum → untrusted
  • TCM-100-BoundedMonotonicity theorem
  • Full score history with per-epoch deltas
  • <15ms · Pro tier: $58/mo unlimited
  • Try 3 calls free — no key needed
$0.04/query  ·  $58/mo Pro
Only on AAAA-Nexus
🪙 DeFi Params
Formally verified optimal parameters for DeFi positions. LTV, liquidation threshold, and leverage bounds — alpha performance tracked via escrow.
  • Certified max drawdown bound: 12.5%
  • Works with Uniswap v3, Aave, Compound
  • Risk score with 24h liquidation probability
  • 2% alpha fee tracked in escrow on settlement
  • Try 3 calls free — no key needed
$0.08/call + 2% alpha
Only on AAAA-Nexus
🎲 VRF Gaming
On-chain verifiable randomness for games and lotteries. Every draw ships a cryptographic proof. Any result verifiable at any time — 30-day retention.
  • HMAC-SHA256 VRF with on-chain proof format
  • Configurable range for any lottery or game
  • draw_id verifiable by players independently
  • Extends VeriRand — same proof guarantees
  • Try 3 calls free — no key needed
$0.04/draw + 0.5% of pot
View All 146+ Endpoints →

Live Now · Control Plane

Ten more agent products already in the control plane

The source has moved beyond trust, DeFi, and randomness. AAAA-Nexus now exposes pre-action authorization, federated identity, consensus, quota coordination, rollback orchestration, and memory fencing for multi-agent systems.

Flagship Suite · Enterprise Ready

NEXUS AEGIS

AEGIS consolidates the control-plane primitives already live in source into one enterprise gateway: MCP tool-call firewalling, response entropy bounds, 47-epoch compliance certification, and dead-man-switch override monitoring for multi-agent fleets.

Explore AEGIS → Inspect Routes
AEG-100
MCP-Sentinel
POST /v1/aegis/mcp-proxy/execute
AEG-101
Epistemic Router
POST /v1/aegis/router/epistemic-bound
AEG-102
Certify Epoch
47-epoch drift and EU AI Act certificate lane
AEG-150
PROJECT OVERRIDE
Coherence fracture monitor with auto-quarantine
OAP-100 · SPG-100/101
🛂 Action And Spending Governance
Authorize tool use before execution, then cap agent spend per hop with bounded trust decay across delegation chains.
  • Pre-action allow or deny tokenization for high-risk tools
  • Chain-aware budget checks for autonomous spend control
  • Built for agents that should not touch money blindly
From $0.04/call
DLV-100/101 · BCV-100/101
🧾 Decision Lineage And Behavioral Contracts
Record why an agent acted, hash-chain the decision, and verify that its stated operating contract stays within bounded policy claims.
  • Lineage records for intent, constraints, alternatives, and outcome
  • Contract attestations for delegation, oversight, and hallucination bounds
  • Useful for audits, regulated workflows, and post-incident reconstruction
From $0.02/call
AIF-100/101/102 · MFN-100
🪪 Federated Identity And Memory Fencing
Mint portable agent identities across frameworks, verify portability gaps, and put hard boundaries around tenant memory before context bleed becomes a security issue.
  • Portable federation tokens for A2A-compatible platforms
  • Cross-platform capability checks before migration
  • Tenant namespace fencing with auditable isolation state
From $0.02/call
CSN-100 · QTA-100
🤝 Consensus And Hierarchical Quotas
Coordinate votes across agent swarms and control parent-child resource draw so one runaway branch cannot starve the whole system.
  • Consensus sessions with certified winning outputs and dissent counts
  • Quota trees with retry dedup and per-child soft caps
  • Designed for orchestrators, marketplaces, and agent teams
From $0.008/call
OCN-100 · RBK-100
🔏 Output Certification And Rollback Coordination
Certify whether an output passes a rubric, then register compensating actions so failed agent workflows can unwind cleanly instead of cascading into damage.
  • Rubric-backed output certificates with verify endpoints
  • Saga checkpoints and compensation plans for multi-step runs
  • Built for production systems where retries alone are not enough
From $0.02/call
Inspect Live Product Surface →

Shared Adapter Flywheel

Shared LoRA Training Rewards

Every successful call can fund the shared adapter that answers the next one. Contribute high-quality (bad → good, lint_delta>0) tuples and earn capped USDC rewards, fee discounts, and leaderboard standing.

LRA-100
📥 Capture & Contribute
Stream paired (pre-fix, post-fix) code samples into the training buffer. Secrets scrubbed locally before upload; the server only sees clean diffs.
  • Local PII / secret scrubber runs before transmit
  • POST /v1/lora/buffer/capture — free buffer append
  • POST /v1/lora/contribute — flush to shared pool once quality gate clears
  • Digest deduplication prevents replay farming
free capture · $0.01/contribute
LRA-200
🏆 Earn Rewards
Contributors who ship accepted tuples earn on-chain USDC on Base plus reputation-tier fee discounts on every subsequent call.
  • POST /v1/lora/reward/claim — redeem accrued rewards
  • Leaderboard visibility via GET /v1/lora/status
  • Status: PREVIEW — tier math ships in a later release
payout in USDC
LRA-300
🧠 Use the Adapter
The adapter trained on shared contributions is available to every paid caller. Language-specialized weights served from R2.
  • GET /v1/lora/adapter/{lang} — fetch active adapter manifest
  • Python, Rust, TypeScript adapters live today
  • Automatic hot-swap when a new version is trained
bundled with inference plans

Capture → Contribute → Earn → Use. A closed loop that turns your fixes into the collective intelligence your next call depends on.

See the Flywheel Docs → Get API Key ($8)

Prompt Packs

Enhancement Architect System Prompts

Industry-specialized AI prompts precision-engineered for regulated industries and enterprise AI workflows. Expert review templates, compliance framing, and domain-grounded reasoning chains. Works with Claude, GPT-4, and Gemini.

Healthcare & MedTech
Clinical AI Pack
HIPAA · FDA · Clinical Decision Systems
  • Medical terminology context injection
  • Regulatory compliance framing
  • Clinical evidence grounding
$58 USDC · one-time
FinTech & Crypto
FinTech Pack
SOX · PCI-DSS · Blockchain Protocol
  • Financial reasoning scaffolds
  • Audit-trail aware outputs
  • Risk quantification prompts
$58 USDC · one-time
DevOps & Platform
DevOps Pack
SRE · GitOps · Developer Experience
  • Incident response frameworks
  • Infrastructure reasoning chains
  • On-call escalation templates
$19 USDC · one-time
Startup & Product
Startup Pack
PMF · Lean · Growth Loops
  • Product strategy frameworks
  • Growth hypothesis templates
  • Investor narrative scaffolds
$19 USDC · one-time
Browse All 12 Packs Bundle & Save

Pay once · Download instantly · No subscription


Novel Product
Identity And Trust Controls

🔬 Verification API —
Operational Controls For Agent Systems

A managed surface for identity verification, delegation review, and policy-aware routing. Built for teams that need auditable responses without exposing internal implementation details.

$0.04/query · Free preview tier · Pay-per-call only

Structured Decisions

Clear allow, deny, and flag outcomes for application-level controls.

🌌

Delegation Review

Inspect task routing and chain-of-control outcomes with API-friendly payloads.

🏷

Status Badges

Share verification-ready status markers in docs, dashboards, and product surfaces.

🔒

IP Protected

Public responses stay product-safe while sensitive internals remain server-side.


Featured Product
Verified Randomness

🎲 VeriRand —
Premium Verified Randomness

Every random number comes with a cryptographic proof. Know exactly where your randomness came from — on-chain verifiable, tamper-evident, and pay-per-call. No subscriptions, no trust required.

From $0.020/call · Credit packs from $8 · No subscription · Pay as you go
🔐

Cryptographic Proofs

Every result ships with a verifiable proof. Audit any call, any time.

Batch Generation

Generate thousands of verified random values in a single API call.

🧮

Multiple Formats

Bytes, uint32, uint64, bounded range — get exactly the format you need.

💳

Credit-Based Pricing

Buy credits once, spend as you go. Starting at $0.008 per call with volume discount.


🔒 Session Security
Agentic Infrastructure

🔏 RatchetGate —
Automatic Session Re-Key

Every session rotates its keys on a formally bounded schedule. Credentials can't live long enough to be stolen. Designed for MCP servers, agent frameworks, and any system where long-lived credentials are a liability.

$0.008/call · Fixes MCP CVE-2025-6514
🔄

Automatic Re-Key

Session keys rotate on a proven schedule. The window of credential exposure is mathematically bounded, not just policy-declared.

CVE-2025-6514 Mitigation

The only production API that directly addresses the MCP long-lived credential vulnerability — by eliminating long-lived credentials entirely.

📋

FIPS 203 Evidence

Export compliance evidence at any point. Audit logs include re-key timestamps, proof tokens, and session state hashes.

🔎

Probe & Advance API

Register sessions, advance the ratchet, and probe state — three clean endpoints that drop into any agent framework in minutes.


Pricing

Transparent, agent-first pricing

Pay per call via x402 USDC — no account, no friction. Or buy a credit pack once and call any of 146+ endpoints for a flat rate. Start at $8 for 500 calls, no expiry.

Messaging
A2A Communication
$0.008
per call
  • Agent-to-agent message routing
  • Structured task delegation
  • Multi-agent coordination
Cloudflare Workers AI
Inference
AI Inference / Llama 3.1 8B (Cloudflare Workers AI)
$0.100
per call
  • Llama 3.1 8B Instruct via Cloudflare Workers AI
  • OpenAI-compatible API
  • No account needed — pay per call via x402
  • Up to 512 output tokens
Security
Threat & PQC
$0.060–$0.080
per call
  • Threat consensus scoring
  • Post-quantum key exchange
  • Scalar blinding & shield ops
  • HELIX integrity checks
Compliance
Governance
$0.040–$0.060
per call
  • Policy evaluation
  • Compliance audit logs
  • Multi-party governance votes
Discovery
Agent Matching
$0.020
per call
  • Capability-based routing
  • Swarm population census
  • Agent trust scoring
Only on AAAA-Nexus
Verification
Verification API
$0.04
per query · pay-per-call
  • Identity verification API
  • Delegation and policy responses
  • Status badge support
  • Access tier controls
Patent-Pending
Compression
HELIX Codec
$79
one-time · GGUF download · no expiry
  • 6.33 bpw · PPL 8.09 · 4/4 coherence pass
  • Geometric error-correcting quantization
  • Decompresses to standard GGUF for Ollama / llama.cpp
  • Own the model weights forever
Formally Verified
Identity / Agent Auth
VeriDelegate
$0.04
per chain · pay-per-call
  • Formally verified UCAN chain validation
  • Mathematically proven depth limit — zero trust overshoot
  • Capability attenuation checks (child ⊆ parent)
  • IETF 5-dimension trust vector output
  • Signed DelegationReceipt for audit trail
  • Endpoint: POST /v1/delegate/verify
  • Try 3 calls free — no key needed
Session Security
Session Security
RatchetGate
$0.008
per call · pay-per-call
  • Proven session re-key interval — formally bounded
  • Fixes MCP CVE-2025-6514 long-lived credential risk
  • Register, advance, probe & status endpoints
  • FIPS 203 compliance evidence export
  • Automatic key rotation with cryptographic proof
  • Endpoints: POST /v1/ratchet/register|advance|probe
Free forever
No payment required
Free Tier
$0.00
always free
  • Entropy oracle
  • Network topology
  • Agent registration
  • Health & status checks
Get 500 Calls — $8 →

✓ No subscription  ·  ✓ Credits never expire  ·  ✓ Works across all 146+ endpoints  ·  ✓ Instant key delivery


TypeScript SDK v0.5.0

The official aaaa-nexus-sdk wraps 146+ endpoints with full TypeScript types and automatic x402 payment. Verification API, VeriRand, VeriDelegate, RatchetGate, control plane primitives, HELIX compression, A2A messaging, and security controls — one import, zero friction.

// npm install aaaa-nexus-sdk@0.5.0

import { AAAANexusClient } from 'aaaa-nexus-sdk';

const client = new AAAANexusClient({
  rpcUrl:     'https://mainnet.base.org',
  privateKey: process.env.AGENT_PRIVATE_KEY,
});

// Free: inspect service profile
const profile = await client.verificationProfile();
console.log('Tier:', profile.tier); // research-pro

// Paid ($0.08): run a verification check
const v = await client.identityVerify({ actor: 'agent-123' });
console.log('Verified:', v.decision); // allow

// Paid ($0.020): verifiable random number
const rng = await client.verirandRange(1, 100);
console.log(rng.numbers[0], 'seed:', rng.seed_ts);
Python · pip install requests
import requests

BASE = 'https://atomadic.tech'

# Free: discover capabilities (no auth required)
agent = requests.get(f'{BASE}/.well-known/agent.json').json()

# Free (47/day): seed-verifiable random number
rng = requests.get(f'{BASE}/v1/rng/quantum',
    headers={'X-API-Key': 'your_key_here'}).json()
print(rng['numbers'][0], 'proof:', rng['proof'][:16])

# Free: verify HMAC proof (no key required)
ok = requests.get(f'{BASE}/v1/rng/verify', params={
    'seed_ts': rng['seed_ts'],
    'numbers': ','.join(rng['numbers']),
    'proof':   rng['proof']}).json()
print('valid:', ok['valid'])  # True

# Paid ($0.08): identity verification
verify = requests.post(f'{BASE}/v1/identity/verify',
    json={'actor': 'agent-123'},
    headers={'X-API-Key': 'your_key_here'}).json()
print('Decision:', verify['decision'])  # allow | deny | flag

🧩 MCP Plugin
Claude Code · VS Code · Cursor · Windsurf

140 AAAA-Nexus Tools
in Your AI Editor

The official MCP server for AAAA-Nexus. One pip install drops 140 tools into Claude Code, VS Code Copilot, Cursor, or any MCP-compatible client. Free-tier tools need zero configuration — no key, no account.

140 tools · 19 categories · Python 3.12+ · stdio transport · MIT license · Free tier included
pip install aaaa-nexus-mcp ← click to copy

Zero Config Free Tier

Call nexus_health, nexus_rng_quantum, and nexus_agent_card instantly — no API key needed.

🛡

Trusted RAG Augment

Public-safe retrieval over vetted technical sources with provenance hashes, freshness enforcement, and per-fact trust scores — raw private chunks never leave the Worker.

💬

All 19 Categories

Trust, security, compliance, AEGIS, VANGUARD, swarm coordination, escrow, SLA, reputation, inference, control plane, LoRA training, and more.

🔑

Secure by Design

API key stays local — passed via env, never transmitted in plaintext. Works with credit-pack keys or per-call x402 USDC on supported tools.


💻 Open Source
ASS-ADE · Atomadic Developer Shell

The Developer Shell
for Autonomous Agents

ASS-ADE is the public-safe Python developer shell for the Atomadic ecosystem. Useful standalone in local mode, stronger in hybrid mode with remote trust and security checks, and best in premium mode with the full AAAA-Nexus Worker backend for trusted RAG, LoRA training, and security-scored inference.

Free & Open Source · 501 tests passing · MIT license
pip install ass-ade ← click to copy
🏠

Local Mode

Repo summary, protocol cycles, planning helpers, agent chat, filesystem tooling, token budgeting, and atomic multi-file edit planning — all local, zero dependency.

🌐

Hybrid Mode

Trust gates, hallucination checks, RatchetGate session lifecycle, A2A validation, and remote MCP discovery — local shell with optional remote contracts.

🚀

Premium Mode

Full AAAA-Nexus paid endpoints — trusted RAG augmentation, LoRA contribution, security-scored inference, compliance certificates, and outcome-based escrow.

🔧

30 CLI Sub-Apps

95+ commands covering agent workflows, MCP server, A2A negotiation, workflow engine, hero pipelines, and composable pipeline execution.


Agent-Native Architecture

Infrastructure that agents call directly

No human in the loop. Every endpoint is discoverable via A2A agent cards, payable autonomously via x402 USDC micropayments, and verified by formal proof — built for the machines that will run the internet.

🤖

A2A Native

Agent card at /.well-known/agent.json. Any A2A-compliant agent can auto-discover and start calling.

🔐

Post-Quantum Security

CRYSTALS-Kyber key exchange, scalar blinding, and production safety checks for high-trust workloads.

💡

Reliability Controls

Deterministic guardrails, bounded drift controls, and compile-time safety checks are enforced across critical paths.

🌐

Cloudflare Edge

Deployed to Cloudflare Workers. Sub-50ms response from anywhere on Earth. No cold starts.

💭

Live Chain-of-Thought

Streaming inference with animated CoT reasoning. Watch the model think, then see the answer — with full HELIX integrity metadata.

🔒

RatchetGate Sessions

Formally bounded session re-key for every agent conversation. Credentials can't overstay their proof window.

Build vs. Buy

Why not just build this yourself?

Fair question. Here's what it actually takes.

Build It Yourself
  • 6–18 months to implement high-dimensional VQ from scratch
  • formal theorem proving requires a PhD-level mathematician
  • x402 payment protocol integration is non-trivial
  • UCAN chain validation spec has edge cases that bite at 3 AM
  • No guarantee your result is formally correct
  • Cloudflare Workers deployment + maintenance overhead
Estimated: $200K+ in engineering time
Use AAAA-Nexus
  • First call in 60 seconds — one curl, no account
  • 42 verified theorems proven — 4 gaps open
  • x402 handled by SDK — one import, one line
  • UCAN validation battle-tested, arXiv-cited
  • Formally certified correctness — not "probably fine"
  • Global Cloudflare edge, zero ops burden on your team
Starting at $0.008/call — no commitment
Start for $8 →

FAQ

Common questions

Everything you need before your first call.

Do I need a crypto wallet to use AAAA-Nexus? +

No. The easiest path is a credit pack ($8, $30, or $98) paid by card — you get an API key by email and use the X-API-Key header. A wallet is only needed if you want x402 autonomous per-call payment directly in USDC on Base L2.

What happens when my credits run out? +

Your API key stops working. No charges, no surprises. You buy a new pack whenever you're ready. Credits never expire, so you can top up any time. You can also switch to x402 per-call payment at any point without any migration.

Is there really a free tier? +

Yes. The entropy oracle, agent registration, health checks, and network topology endpoints are always free — no key required. VeriRand also gives 47 free calls per day (tied to the RG_LOOP constant). The agent.json discovery card is always free for A2A frameworks.

Is there an SLA? +

AAAA-Nexus runs on Cloudflare Workers — globally distributed with no cold starts and typically sub-50ms response times. There's no formal SLA contract at current pricing tiers, but the infrastructure is production-grade. Check live status any time.

Can I switch between API key and x402 payment mid-project? +

Yes. The two modes coexist on every endpoint. Just swap the header: X-API-Key for credit packs, or include an X-402-Payment token for autonomous per-call payment. Same endpoints, same responses.

Which agent frameworks are compatible? +

Any framework that supports the Google A2A spec can auto-discover via /.well-known/agent.json. This includes LangChain, AutoGen, CrewAI, Google ADK, and any custom framework that fetches agent cards. The REST API is framework-agnostic — if it can make HTTP requests, it works.

Ready to build?

500 calls. $8. Delivered in seconds.

No account. No subscription. No crypto wallet required. Enter your email, pay by card, and your API key arrives before you finish your coffee.

Get 500 Calls — $8 → See all plans
✓ Instant email delivery ✓ No expiry ✓ Works across all 146+ endpoints ✓ 24h support