Apache-2.0 — Open Source

The 100x
AI Engineering
Harness

A sovereign, deterministic AI coding harness with neural context intelligence, dynamic agent swarms, adversarial verification gates, and a 7-layer provenance memory system — for engineering at scale.

0 Specialized Agents
0 Validated Skills
0 End-to-End Workflows
0 Language Ecosystems
aether-swarm — zsh
$ _
Lease Acquired
src/auth/middleware.ts
6 Gate Passed
Adversarial Verification
Memory Provenance
SHA-256 Merkle Root

Built for deterministic
engineering intelligence

Every module in AETHER is purpose-built. From neural context slicing to adversarial falsification gates — no hollow abstractions, every engine is deterministic and verified.

Neural Context Intelligence

Personalized PageRank over AST symbol graphs ranks every file by relevance. 0/1 Knapsack allocates tokens — full source, skeleton, or evicted.

src/core/contextEngine.ts

Dynamic Agent Swarm

TCCI score routes tasks to 1 engineer or a DAG swarm. Exclusive file-write lease locks prevent parallel edit collisions across 117 specialists.

src/core/swarmEngine.ts

Adversarial Verification Gate

Six hostile challengers — Architecture, Code Breaker, Security Attacker, Performance Critic, Regression Investigator, Judge — require executable reproducers.

src/core/adversarialEngine.ts

7-Layer SHA-256 Memory

Provenance-bound memory from Session through Verified Knowledge to Failure Memory. Auto STALE_REVALIDATION_REQUIRED on file mutation.

src/core/memoryEngine.ts

Ten deterministic engines

Each engine is a self-contained, testable module — no cross-contamination, no magic. Together they form the complete AETHER reasoning pipeline.

Context

Context Intelligence

AST-aware symbol graph ranking with Personalized PageRank and token budget knapsack compression.

Swarm

Swarm Orchestration

Dynamic DAG swarm topology with Task Complexity & Coupling Index, exclusive file-lease locks, and conflict resolution.

Predictive

Predictive Impact

Deterministic AST call-graph traversal + git co-change coupling with automatic Targeted Verification Plan synthesis.

Adversarial

Adversarial Engine

Six hostile challengers testing architecture, security, performance, concurrency, and regression boundaries with executable reproducers.

Security

Security Suite

Shannon entropy secret scanning, prompt-injection spotlighting, MCP tool manifest guard, egress policy, and integrity-chained audit log.

Memory

Memory Engine

7-layer store bound to SHA-256 file hashes with automatic staleness downgrade on mutation. Session through Verified Knowledge.

Debug

Debug Engine

Scientific hypothesis debugging state machine. Hard gate requiring non-zero failing reproducer exit code and Bayesian falsification before minimal fix.

Benchmark

Benchmark Engine

Multi-trial benchmarking with mean, standard deviation, 95% CI, and regression tracking across all 711 components.

Audit

Audit Engine

Continuous quality scoring across 5 deterministic rubrics. Auto-refines weak specs and re-tests to 99.6+/100 quality gate.

Self-Improve

Self-Improvement Engine

Closed-loop test-score-refine cycle. Evaluates all agents and skills deterministically, auto-refines weak specifications to target score.

Registry

Ecosystem Registry 2.0

Dynamic agent creation, skill composition pipelines, MCP server lifecycle, SemVer plugin resolution with topological sort & cycle detection.

Every role covered.

From Principal Systems Architect to Security Attacker — each agent has a strict role, scope, responsibility list, handoff rules, and verification gate.

Architecture (8)

systems-architect-prime
Principal Blast-Radius Architect
api-contract-architect
Interface Contract Designer
data-model-architect
Database Schema Evolution
microservice-boundary-architect
Service Mesh Partitioning
performance-architect
Throughput & Latency Modeling
observability-architect
Telemetry & Tracing Design
sre-platform-architect
Reliability Engineering
auth-security-architect
IAM & Zero-Trust Design

Implementation (24)

domain-implementer-ts
TypeScript Core Engineer
domain-implementer-py
Python Core Engineer
domain-implementer-go
Go Core Engineer
domain-implementer-rust
Rust Systems Engineer
frontend-crafter
React / UI Engineering
db-migration-surgeon
Schema Migration & Data
infra-provisioner
Terraform / OpenTofu IaC
api-gateway-engineer
API Gateway & Routing

Adversarial (6)

adversarial-arch-challenger
Architecture Falsification
adversarial-code-breaker
Edge Case & Logic Breaking
adversarial-security-attacker
Injection & Auth Boundary Testing
adversarial-perf-critic
Algorithmic Complexity Attack
adversarial-regression-investigator
PASS_TO_PASS Baseline Breaker
adversarial-judge
Final Consensus Verdict

Debugging (4)

tdd-verification-engineer
Red-Green-Refactor Gate
build-failure-localizer
Compiler Error Root Cause
runtime-debugger
Multi-Hop Runtime Analysis
test-flakiness-hunter
Flaky Test Detection & Fix

Security (5)

secret-scanner
Shannon Entropy Detection
injection-guardian
Prompt Injection Defense
auth-boundary-enforcer
IAM & OAuth2 Validation
supply-chain-auditor
Dependency & License Audit
vps-hardening-specialist
WireGuard, nftables, systemd

Memory & Verification (4)

memory-curator
SHA-256 Provenance Memory
continuity-engineer
Merkle Session Checkpoints
independent-reviewer
Final Quality Gate
benchmark-verifier
Regression Benchmark Runner

Lazy-loaded domain packs

Skills activate via 28-token AST predicates. Zero tokens consumed until the activation predicate matches the active file graph.

Design Craft

105
Design Skills
OKLCHAPCAWCAG AAASubgridThree.jsD3.js

Polyglot Core

332
Core Engineering Skills
TypeScriptPythonGoRustSQL

Cloud / AWS / VPS

85
Infrastructure Skills
EKSIAMWireGuardPatroniCaddy

Video / Media NLE

18
Media Editing Skills
FFmpegEBU R128WhisperHLSAV1

Every skill carries a deterministic activation predicate, required information list, execution steps, useful tools, expected outputs, verification criteria, and failure-case handling — all defined in structured frontmatter. Skills are verified against AETHER's 0-filler quality gate before entering the registry.

Domain packs (core-architecture, distributed-systems, cloud-native, etc.) are loaded on demand. Installing 500 skills adds 0 tokens to the base prompt until an activation predicate matches the active AST/file graph.

DynamicEcosystemRegistry.composeSkills() chains existing skills into higher-order composite pipelines with deterministic scoring, combined trigger keywords, and required permission enforcement.

Intelligence layered by domain

Each program composes existing AETHER engines rather than reimplementing them — adding the specialized reasoning layer each domain demands.

P01

Development Intelligence

Research-first TDD gate with deterministic AST skill discovery, Red-Green-Refactor enforcement, and debugging hypothesis ledger.

Informed by ECC & Superpowers
P02

Reverse Intelligence

Isolated, permission-controlled reverse-engineering. Application architecture extraction, dependency discovery, and automated technical reporting.

Informed by REA (morluto)
P03

Autonomous Execution

Goal-to-plan resumable execution layer. Dependency-structured action plans with deterministic verification predicates per step and crash recovery.

Informed by OpenHands & OpenManus
P05

Browser Intelligence

Unified browser engine with A11y ref snapshots, deterministic CSS/ARIA DOM locators, and vision coordinate fallback — ranked by determinism.

Informed by Playwright, Agent Browser, Browser Use
P06

Visual Computer Intelligence

Perception → Reasoning → Action → Verification loop. OmniParser-style UI element grounding with confidence-floor abort guarantees.

Informed by OmniParser & Skyvern
P07

Browser Runtime

Provider abstraction for Playwright cross-browser, Lightpanda CDP (BSL-1.1, out-of-process only), and bare Chromium backends.

Informed by Playwright & Lightpanda
P09

Creative Visual Intelligence

ComfyUI-style node graph (clean-room data model) + Three.js-style 3D scene graph. Topological scheduling with cycle detection.

Informed by ComfyUI (out-of-process) & Three.js
P10

Crypto & Integrity

SHA-256 content-addressed hashing, Merkle tree continuity checkpoints, FNV-1a fast hashing, and integrity-chained audit logs.

Standard cryptographic primitives
MCP

Model Context Protocol

Full JSON-RPC 2.0 MCP server engine. 8 AETHER tools, capability negotiation, permission enforcement, and structured error codes.

Anthropic MCP Specification

Install anywhere. Run everywhere.

The selective installer deploys agents, skills, hooks, and workflows to 5 major AI coding harnesses — and the MCP server. Under 650 idle tokens.

Single Command Install

Install everything across all supported harnesses with one command.

$ npx tsx bin/aether.ts install --all --target all-harnesses --mcp

Verified execution pipelines

Every workflow is a 5-stage pipeline with explicit dependencies, failure handling, and completion criteria.

1
Neural Context
Slice context to <25% capacity. Identify AST blast radius.
2
Red-Green Contract
Write failing tests. Verify RED phase before implementation.
3
Leased Swarm
Exclusive file-write leases. Minimal implementation. Local verify.
4
Adversarial Gate
6 hostile challengers. Executable reproducers required. Zero blockers.
5
Provenance Commit
SHA-256 memory record. Merkle checkpoint. Zero-loss continuity.

Provider-agnostic intelligence

AETHER's harness layer abstracts every major AI provider through a unified interface. Swap models without rewriting agents, skills, or workflows.

Anthropic
Primary Provider
Claude Opus Claude Sonnet Claude Haiku claude-fable-5-2

Native first-class support via Anthropic SDK. Full extended thinking, tool use, vision, and the 1M context window. Default for AETHER agents and all Fable-tier workloads.

Google
Cloud Provider
Gemini 3 Pro Gemini 2.5 Flash Gemma 3

Vertex AI and Generative Language API integrations. Supports multimodal reasoning, tool calling, and function-calling schemas with adapter parity.

OpenAI
Cloud Provider
GPT-4o o3 o4-mini

OpenAI API adapter for GPT-4o, reasoning models (o3, o4-mini). Structured outputs and function-calling interface through the OpenAI SDK wrapper.

Local / Open
Self-Hosted
Ollama vLLM Llama 4 DeepSeek Qwen Mistral

Fully offline via Ollama or vLLM inference servers. Supports Llama 4, DeepSeek, Qwen, Mistral, and any OpenAI-compatible endpoint. No data leaves your network.

Sub-35ms safety gates

Four lifecycle hooks intercept every prompt, tool use, edit, and compaction with non-blocking execution budgets.

pre-prompt-security-and-slicePrePromptSubmit

Scan untrusted files for indirect prompt injection, strip high-entropy secrets, compute AST PageRank context slice.

Priority: 10Budget: 50ms
pre-tool-command-sandboxPreToolUse

Verify exclusive file-write lease held. Block destructive shell AST patterns. Bash, Shell, Edit, Write matchers.

Priority: 20Budget: 35ms
post-edit-incremental-verifyPostToolUse

Invalidate stale Layer-3 memory hashes for modified files. Trigger targeted AST-mapped unit test on every edit.

Priority: 30Budget: 1200ms
pre-compact-merkle-checkpointPreCompact

Persist Zero-Loss Continuity Merkle checkpoint with falsified debugging hypotheses and verified DAG task state.

Priority: 40Budget: 80ms

15 integrated ecosystems

License-audited adapters with strict isolation boundaries. Permissive ecosystems get native in-process adapters; copyleft projects are sandboxed to out-of-process HTTP/CDP only.

Every boundary enforced

Security is structural. From untrusted content spotlighting to license boundary enforcement, AETHER's threat model is explicit.

Untrusted Content Spotlighting

All repo files, web pages, and MCP payloads pass through the spotlighting gate. Injection directives are quarantined inline — never reach the model as live instructions.

MCP Tool Manifest Guard

Untrusted tool registration is impossible. Only tools in the local AETHER_MCP_TOOLS manifest are accepted. Permission denials emit structured JSON-RPC -32001 errors.

License Boundary Enforcement

Permissive (MIT, Apache-2.0) ecosystems get native adapters. Copyleft (AGPL-3.0, GPL-3.0, BSL-1.1) are strictly out-of-process HTTP/CDP only.

Shannon Entropy Secret Scanning

Detects high-entropy tokens (Shannon > 4.35) matching credential signatures. Zero hardcoded secrets policy enforced on every commit.

Network Egress Policy

Shell commands are gated by configurable egress rules. Only whitelisted hosts reachable. Enforced at the tool-call boundary.

Integrity-Chained Audit Log

Every gate decision is recorded with SHA-256 hash chain. Tampering detectable. Complete forensic trail from security scan through adversarial verdict.

Built on permissive foundations

Every dependency, inspiration, and ecosystem is credited with SPDX license identifiers. No uncredited borrowings.

Ready to build with
deterministic intelligence?

Open source. Apache-2.0. 117 agents. 540 skills. Zero filler.