aeqnt

Artificial Entities, Quantized

artificial · entities · quantized · neural · topology

Describe what you need. Aeqnt finds, builds, and composes the right entity for the job. Every request makes the system smarter.

Limited Experiment

Aeqnt is in a limited experiment mode. We're testing composable AI entities with a small group of early participants.

Join the Experiment →

Bring Your Own Keys Multi-Provider Self-Evolving

The Stack

Three composable primitives

Every entity is assembled from the same clean hierarchy. Skills are shared. Tools are swappable. Agents orchestrate.

Agents

Who
executes

Curated bundles of skills with identity, system prompts, and orchestration logic. Route tasks. Decompose complexity. Deliver results.

Skills

What
they know

Portable domain expertise. Phishing analysis, threat modeling, data pipelines. Shared between agents. The composable atom of intelligence.

Tools

How
they work

Web search, API calls, file I/O. Skills declare what they need. Swap implementations without touching domain logic. The plumbing stays invisible.

∞

Skills

∞

Agents

∞

Compositions

Limits

Routing Cascade

Ask once. We figure it out.

Your request flows through a priority cascade. Existing entities are matched first. New ones emerge only when needed.

01 Skill match — existing agent has the right skill

02 Template deploy — pre-built blueprint, instant setup

03 Deep analysis — LLM reasoning, optimal routing

04 Compose new — custom entity, tested and deployed live

The result

Every request makes it smarter

Entities are indexed for future reuse. Skills accumulate. The system learns your patterns. Over time, faster matches, better compositions, sharper results.

Execution

From request to result

Four phases. Fully autonomous. Human-in-the-loop when it matters.

Understand

Parse intent, decompose compound tasks, assess complexity

Match

Vector similarity + LLM re-ranking across all agents and skills

Execute

Tool-use loops, parallel sub-agents, real-time streaming

Evolve

Feedback drives versioned improvements with rollback

Open Protocols

No walled gardens

Open standards. Connect to any ecosystem. Export any entity and run it anywhere.

MCP

Model Context Protocol

Connect to any MCP server. Expose your entities as MCP tools. Bidirectional integration with the broader AI ecosystem.

A2A

Agent-to-Agent

Federate with external entities. Discovery via agent cards. Cross-platform task delegation and collaborative execution.

BYOK

Your Keys, Your Rules

Bring your own API keys. Anthropic, OpenAI, Google, xAI, Azure, and 6 more providers. No middleman, no markup. You own the intelligence.

Security First

Your data stays yours

Enterprise-grade security without enterprise complexity.

AES-256

Encrypted at Rest

API keys encrypted with AES-256 (Fernet). Passwords hashed with bcrypt. Session tokens signed with HMAC-SHA256.

Isolated

Per-User Isolation

Every user gets a private workspace. Multi-tenant architecture with strict data boundaries. No cross-user data leakage.

Stateless

Safe Sharing

Share entities via MCP and A2A protocols. Public callers run in stateless mode — no access to your workspace, files, or keys.

Sandbox

Sandboxed Execution

Code runs in restricted sandboxes. AST-validated expressions. Rate limiting on all endpoints. Content policy enforcement.