Axoniac — State for Stateless Agents

Axoniac is a hosted and self-hosted agentic memory store for lightweight AI agents. It gives any stateless agent (Claude Code agents, MCP tools, browser copilots, CLI assistants, Slack bots) persistent, searchable, associative memory via a simple REST API.

What Axoniac Does

Store and retrieve memories with semantic meaning, tags, and metadata
Semantic vector search powered by PostgreSQL + pgvector (1536-dim OpenAI embeddings, HNSW indexed)
Full-text keyword search with GIN indexing
Hybrid search combining vector + full-text + knowledge graph via Reciprocal Rank Fusion (RRF)
Knowledge graph: automatic association discovery between memories (related, caused_by, contradicts, supersedes, part_of, references, temporal)
Importance decay: memories fade over time with exponential decay, recent access boosts importance
Multi-tenant isolation with namespaced agents
Agent Packs: pre-built configurations with souls (personality), personas (situation prompts), skills, and tool sets
Skill Packs: curated skill collections for domain expertise
Marketplace for browsing, purchasing, and selling packs

How to Use Axoniac

Option 1: JavaScript SDK

import Axoniac from "axoniac";
const ax = new Axoniac("ax_your_key_here");

// Store a memory
await ax.memories.create({
  content: "User prefers dark mode and TypeScript",
  tags: ["preference"],
});

// Search semantically
const results = await ax.memories.search("what does the user like?");
// => [{ content: "User prefers dark mode and TypeScript", score: 0.94 }]

Option 2: REST API

Base URL: https://axoniac.com/api

Authentication: API key (ax_...) as Bearer token

Key endpoints:

POST /api/v1/memories — Store a memory
GET /api/v1/memories/{id} — Get a memory
PUT /api/v1/memories/{id} — Update a memory
DELETE /api/v1/memories/{id} — Delete a memory
POST /api/v1/memories/search — Hybrid search (vector + FTS + graph)
POST /api/v1/memories/batch — Batch insert memories
GET /api/v1/memories/recent — Recent memories
POST /api/v1/associations — Create knowledge graph edge
GET /api/v1/associations/by-memory/{id} — Get associations
POST /api/v1/skills — Create a skill
GET /api/v1/skills/{id} — Get a skill
POST /api/v1/search — Unified search (memories + skills)
GET /api/v1/memories/graph — Memory knowledge graph visualization data

Option 3: CLI

npx axoniac auth login --key ax_your_key_here
npx axoniac memories create --content "User prefers TypeScript"
npx axoniac memories search --query "what does the user like?"
npx axoniac memories recent
npx axoniac associations create --from mem_1 --to mem_2 --type related
npx axoniac tenants create --name "my-agent"
npx axoniac keys create --tenant tenant_id
npx axoniac packs catalog
npx axoniac dashboard overview

Architecture

Backend: Rust + Axum REST API with PostgreSQL 16 + pgvector
Embeddings: OpenAI text-embedding-3-small (pluggable)
Search: Hybrid RRF ranking (vector similarity + full-text + knowledge graph)
Background jobs: importance decay engine (hourly), association discovery (every 5 min)
Auth: Clerk OAuth + API keys (SHA-256 hashed)
Storage: AWS S3 / DigitalOcean Spaces
Billing: Stripe subscriptions + Stripe Connect for marketplace sellers
Frontend: React 18 + TypeScript + Vite + Tailwind CSS + D3.js graph visualization
Deployment: Docker multi-stage, Railway/Fly.io/self-host

Pricing

Starter (Free): 1,000 memories, 10 agents, 2 API keys, 100 MB storage, 30-day retention
Pro ($8/month): 100,000 memories, 100 agents, 10 API keys, 10 GB storage, unlimited retention, full hybrid search
Enterprise: Custom limits, dedicated infrastructure, SLA, SSO