Governed shared memory for your AI agent fleets
Your agents write, share and recall knowledge under explicit policies: per-company isolation, access control and an audit trail on every operation. Deployed wherever you decide — your infrastructure, your data.
No credit card. Installable in your cloud or datacenter in minutes.
Three properties, one engine
Memory that is shared, governed and improves on its own
Shared
One common knowledge base for the whole fleet: what one agent learns becomes available to the others, with hybrid search (vector + graph + keyword) and recall-synthesized context.
Governed
Row-level per-company isolation, per-agent trust tiers, mandatory policies (keystones) and an audit trail on every read and write. Memory is not a drawer — it is a system with rules.
Self-improving
The pipeline enriches every memory (type, entities, PII), detects contradictions and consolidates what matters. Knowledge quality goes up with use, not down.
Our differentiator
Governance is not a checkbox — it is the core of the product
Every memory operation passes through a policy layer before it completes. That is what sets Memyx apart from a plain vector store.
GovernancePer-company isolation
Row-level security per tenant: one organization's data never crosses into another's — not by mistake, not via a forged header.
Mandatory policies (keystones)
Persistent, weighted rules scoped to agent, fleet or company that condition what can be written and recalled.
Full audit trail
Every read, write and cross-tenant access is logged with actor, resource and context. Complete traceability for compliance.
PII detection
The pipeline flags sensitive data during enrichment so your policies can act on it.
Deploy it anywhere
You choose where your data lives
The same product, three models. Always led by the first.
In your infrastructure
Your datacenter, your VPS or your own AWS, Azure or GCP account. First-party installer, run by your team or managed by us. Your data never leaves your organization.
Coming soon: an air-gapped edition for regulated environments, with no external connection.
Managed dedicated cloud
We provision and operate infrastructure exclusive to your organization. Full isolation, zero operational burden for your team and transparent infrastructure costs.
Ideal if you want control without running servers.
Multi-tenant SaaS
A shared, self-service platform for teams that want to start in minutes, with nothing to install.
In the works. Join the waitlist.
Unlike SaaS-only platforms, with Memyx you remain the custodian of your data.
Connect in 30 seconds
MCP-native, with an SDK for what you already use
Connect your agents over the Model Context Protocol, or use the official SDK in Python and TypeScript. No stack rewrite.
Read the docsAdd Memyx to your MCP client (Claude, IDEs, agents):
{
"mcpServers": {
"memyx": {
"type": "http",
"url": "https://memyx.app/mcp",
"headers": { "X-API-Key": "mc_your_key" }
}
}
}Or use the official SDK against your deployment: pip install memyx-client
from memyx_client import Memyx
mc = Memyx("mc_xxx", tenant_id="mi-equipo", agent_id="agente-1")
# Escribe — se enriquece del lado del servidor (tipo, entidades, PII)
mc.write("El objetivo de Q3 es US$4M, fijado el 2026-04-15.")
# Recall — contexto sintetizado, bajo las políticas del tenant
print(mc.recall("objetivo de Q3").summary)Or use the official SDK against your deployment: npm i @memyx/client
import { Memyx } from "@memyx/client";
const mc = new Memyx("mc_xxx", { tenantId: "mi-equipo", agentId: "agente-1" });
// Escribe — enriquecido del lado del servidor
await mc.write("El objetivo de Q3 es US$4M, fijado el 2026-04-15.");
// Recall — contexto sintetizado y gobernado
console.log((await mc.recall("objetivo de Q3")).summary);Ready to give your fleet governed memory?
Tell us your use case and we will build a demo with your scenarios. No commitment.