Every conversation, your AI meets your users for the first time. AmnesiaIQ is the invisible memory layer that makes it stop — persistent context across sessions, any LLM, any product.
Every session, your AI introduces itself to the same user it spoke with yesterday. They re-explain their preferences, their history, their context — every time. That's not intelligence. That's amnesia.
AmnesiaIQ runs silently inside your product. Your users never know it exists — the AI just knows them. No memory UI to manage. No settings to configure. No 50 First Dates.
Drop a domain pack in your packs directory. Set your Postgres and Redis URLs. Start the server. amnesia-saas handles the rest.
When a user signs up, your platform calls POST /v1/tenants. AmnesiaIQ provisions their memory store invisibly. No user action required.
Every conversation, the AI recalls what it knows and stores what it learns. Users experience an AI that genuinely knows them. They never know why.
Domain packs extend AmnesiaIQ with vocabulary and decay rules specific to your vertical. One Python file. Drop it in. Your AI immediately knows the language of your business.
Remembers each trader's risk rules, preferred instruments, active strategies, and position context — without them ever repeating themselves.
Knows each client's entity structure, identified deductions, prior strategy discussions, and outstanding obligations — across every session.
Tracks player style, story progression, behavioral patterns, and session context — personalizing every interaction without the player configuring anything.
Any vertical. One Python file. See the SDK →
AmnesiaIQ runs locally for personal use or at scale for your SaaS customers. Same memory engine, same domain packs, different infrastructure.
Local-first, single-user deployment. Runs as an MCP server via stdio. Perfect for OpenClaw, Claude Desktop, and personal AI workflows.
Multi-tenant deployment for products with real customers. HTTP/SSE MCP transport. Each customer gets isolated memory without knowing it exists.
/v1/mcpThe pack SDK is open source. Define categories, set decay windows, validate with one command. The memory engine handles the rest.
# packs/my_domain.py from amnesia.registry import register_domain register_domain( domain="trading", description="Options and equities context", categories=[ "strategy", # recurring playbooks "risk_rule", # non-negotiable limits "instrument", # actively traded "position", # current exposure ], decay_windows={ "position": 7, # days "strategy": 365, "risk_rule": 730, # near-permanent }, )
AmnesiaIQ is in active deployment across NeoSwarm and SwarmIQ. Request access to the full engine or explore the open source SDK.
We'll reach out directly. No spam.