AEN
Open-Source Distributed AI Execution Platform
Product Overview
AEN is the first open-source, community-powered AI execution platform. It transforms idle consumer and enterprise compute — laptops, workstations, local GPU endpoints, and cloud APIs — into a secure, verifiable, distributed AI execution fabric. Instead of running AI workflows on a single machine, AEN lets organizations, developer communities, and teams pool their compute, share AI capabilities, and deploy intelligent workflows across a coordinated network of Hosts and Workers.
The Market Gap
Today's AI ecosystems are fragmented. Organizations repeatedly rebuild the same workflows, valuable local compute sits idle while developers pay premium cloud prices, and AI capabilities remain locked inside individual applications with no way to share or distribute them across a community.
Systems Strategy
Built an open-source distributed AI execution platform with an Ed25519 cryptographic identity layer, lease-based DAG scheduling, adaptive verification, and a provider-agnostic architecture that scales from a single laptop to hundreds of distributed workers.
Open-Source Repository
This project is fully open-source under the MIT license. Clone the repository and run it locally in under five minutes. No mandatory cloud dependency, no vendor lock-in.
git clone https://github.com/keyng-david/AEN.git
cd AEN
pnpm install
pnpm devProduct Features
Distributed Execution
- DAG Orchestration: Requests decomposed into Directed Acyclic Graphs with each step leased to a verified worker.
- Lease-Based Scheduling: Renewable leases with TTL; expired leases automatically return work to the queue.
- Capability-Aware Routing: Tasks matched to workers by verified capabilities, models, VRAM, GPU type, and load.
- Speculative Execution: Slow tasks re-dispatched to additional workers to prevent bottlenecks.
- Core Agent Runtime: Think to Tool to Observe to Reflect to Retry to Verify to Finish reasoning loop per lease.
Adaptive Verification
- Spot-Check Verification: Configurable percentage of completed steps re-run on a second worker.
- Consensus-3 Execution: Critical workflows run on three workers simultaneously; divergence triggers quarantine.
- Full Replication: Highest-trust mode where every step is independently executed and compared.
- Quarantine System: Workers that fail verification are quarantined for investigation and recovery.
Community Networks
- Community AI Networks: Private or public compute pools with shared policy, economy, and verification rules.
- Host / Worker / Hybrid Modes: One installation, three runtime roles for flexible deployment.
- Community Economy: Credits, access policies, contribution rewards, and token-scoped access grants.
- Real User Connectors: Telegram (fully integrated), Web Chat UI, and extensible framework for Discord, Slack, WhatsApp.
- Reusable Skills & MCPs: Communities publish approved capabilities that workers inherit automatically.
Security & Infrastructure
- Ed25519 Cryptographic Identity: Every node gets a keypair with challenge-response authentication.
- Signed Capability Manifests: Workers submit cryptographically signed declarations of providers, models, and hardware.
- Least-Privilege Data Scoping: Workers receive only the context required for their current lease.
- Provider-Agnostic: Ollama, LM Studio, Fireworks AI, OpenAI, OpenRouter, AMD Developer Cloud, and any OpenAI-compatible endpoint.
- Cloudflare Quick Tunnel: One-click public endpoint for remote workers with no account needed.
- Event-Driven Observability: Every state transition emits an event for dashboards, metrics, logs, and auditing.
Server Requirements
- Node.js 20+ runtime
- pnpm (recommended) or npm package manager
- SQLite for local-first deployment (default)
- PostgreSQL for production deployments (optional)
Required API Keys
- Ollama or LM Studio for local inference (optional)
- OpenAI-compatible API key for hosted inference (Fireworks AI, OpenRouter, etc.)
- Telegram Bot Token for connector integration (optional)
Repository Access
Fully open-source. No cost, no vendor lock-in, no mandatory cloud dependency.
Technical Specifications
Need Custom Build?
Need a feature this doesn't have, or a system built from scratch? Book a 20-minute call and we'll scope it.