AppliedAgentics / Memory Core / Patent AU 2026904317

Graph-native persistent memory
for production AI agents.

Your agents start cold. Memory Core™ ends that. Persistent, graph-structured memory that compounds across sessions, across agents, across teams — hosted on Azure, MACC-eligible, zero infrastructure overhead.

Talk to us about access →

Azure Marketplace MACC-eligible — purchases count 100% against your Azure committed spend Azure AI Foundry · Copilot Studio integration MCP-native

Patent

AU 2026904317

Architecture protected at claim level — graph-native persistent memory substrate.

Azure Marketplace

MACC-aligned · transactable

Subscription + metered usage count 100% against Microsoft Azure Consumption Commitment.

Data plane

Per-tenant · enforced

Tenant isolation enforced at the data plane, not contractual. No cross-tenant queries possible by design.

Publisher

AppliedAgentics Pty Ltd

ABN 98 697 110 517 · Australia · Verified Azure Marketplace publisher.

The problem

Every session starts cold. Every agent starts blank.

Enterprise teams building on Azure AI Foundry, Microsoft 365 Copilot Studio, or any LLM framework hit the same structural wall.

Session isolation

Agents forget when the session ends.

No matter what they learned, resolved, or reasoned through — the next call starts from zero. Institutional knowledge dies at context window boundaries.

Multi-agent context loss

Pipelines don't share what they know.

Agent A completes a task. Agent B starts the next step with no memory of what Agent A established. Every handoff is a cold start.

Knowledge that doesn't compound

The tenth run is as blind as the first.

Without a memory substrate, agents can't improve from experience. Every deployment cycle restarts the learning curve.

No structured retrieval

Vector search finds similar — not related.

Vector-only systems retrieve semantically similar content. Graph-native retrieval traverses actual relationships — cause, dependency, sequence, identity.

What Memory Core is

A substrate, not a feature.

Memory Core operates below the agent layer — a persistent, graph-native knowledge substrate that any agent, any framework, any session can write to and read from.

Graph-native, not vector-only

Relationships are first-class. Memory Core stores how concepts connect — not just that they're similar. Agents can traverse dependencies, timelines, identities, and causal chains, not just find nearby embeddings.

Persistent across sessions

Memory survives session end. What an agent established in yesterday's run is available in today's. Context accumulates rather than resets. The memory substrate outlives any individual agent invocation.

Multi-agent and cross-team

One memory layer, many agents. Agents in the same pipeline share a substrate. Teams can maintain isolated tenant namespaces within the same instance. No cross-tenant leakage by design — enforced at the data plane, not just contractual.

Memory reinforcement

The substrate improves with use. The longer your agents run on Memory Core, the more relevant the context they surface. Production deployments outperform fresh starts on the same task — without retraining and without manual curation.

Framework-agnostic

REST API and MCP server endpoints — Memory Core works with any LLM framework, any agent orchestrator. Native integration with Azure AI Foundry and Microsoft 365 Copilot Studio; also compatible with LangChain, AutoGen, custom orchestration layers.

Hosted SaaS — nothing to deploy

No infrastructure to manage. No vector database to provision, tune, or back up. Memory Core is a managed substrate — subscription activates via Azure Marketplace and tenants provision automatically. Your team writes agents, not memory infrastructure.

// Memory Core MCP tool surface — compatible with Azure AI Foundry + Copilot Studio POST /memory_store → { node_id, links_created, graph_updated } GET /memory_search → { nodes[], relationships[], confidence } GET /memory_boot → { role_context, prior_session_summary, active_nodes } POST /memory_checkpoint → { session_id, nodes_written, graph_delta }

Azure integration

Native across the Microsoft AI stack.

Memory Core is purpose-built for Azure-first AI deployments — from developer evaluation through enterprise production.

MACC · Committed spend

Counts against your Azure commitment

Memory Core is a transactable Azure Marketplace offer. Every dollar spent — subscription and metered usage — counts 100% against your organisation's Microsoft Azure Consumption Commitment. For organisations with existing MACC balance, Memory Core is accessible from budget already committed. No new purchase order required.

Azure AI Foundry · Discovery

Appears where agents get built

Memory Core registers as an MCP server in Azure AI Foundry's agent-build toolchain. Developers building Foundry-native agents discover and configure Memory Core at the point of building — not through a separate procurement motion. Bottom-up adoption at developer evaluation; top-down procurement via MACC at team scale.

Copilot Studio · Enterprise agents

Memory provider for enterprise agents

In Microsoft 365 Copilot Studio, Memory Core appears as a memory provider option when building Copilot agents. Enterprise teams extending Copilot with custom agents get persistent, structured memory without building a memory infrastructure. The same substrate that powers custom agent pipelines powers Copilot extensions.

Access

Memory Core is shipping. We're talking to partners ahead of Marketplace transact.

The IP is filed. The infrastructure is live on Azure. The substrate is in production today. We're working with a small group of platform partners and enterprise design partners ahead of public Marketplace transact — and we'd like the conversation with the right teams.

Platform partners

Hyperscaler conversations — Microsoft, AWS, Google Cloud — about Memory Core as the graph-native memory substrate beneath partner-managed AI platforms.

Frontier labs

Research and product teams at frontier labs exploring persistent memory architectures for agent systems. Open to research partnership conversations and joint evaluation.

Enterprise design partners

Enterprise Azure teams running production AI agents and looking for graph-native, data-plane-isolated memory under MACC. Private offer terms, dedicated support during preview.

Talk to us about access →

Pricing is being shaped against design-partner usage. Public Marketplace pricing will publish when we are confident the structure reflects the value delivered. Until then, every conversation is bespoke.

Differentiation

Not vector search. A memory substrate.

Azure AI Foundry Agent Service includes vector-based session memory during preview. Memory Core is a different class of product.

Capability	Foundry Agent Service Memory	Memory Core
Memory type	Vector-only	Graph-native + vector
Cross-session persistence	Session-scoped	Persistent substrate
Multi-agent shared memory	Pipeline-scoped only	Cross-agent, substrate-level
Works outside Foundry	Foundry-locked	Any LLM framework
Multi-team isolation	—	Per-tenant, data-plane enforced
Compounds over time	—	Memory that compounds with use
Relationship traversal	Semantic similarity only	Graph-native — cause, sequence, identity
Commercial status	Free during preview · in-product memory	Private preview · partner conversations open · Marketplace transact pending

Compliance detail

Built for enterprise security requirements.

Memory Core is designed for production AI deployments in regulated and security-sensitive environments. The top-of-page strip carries the scan-signals; this section expands the specifics.

Encryption

At rest + in transit

All data encrypted at rest. All API traffic TLS 1.2+. Per-tenant encryption key isolation available at Enterprise tier.

Data isolation

Data-plane enforced

No cross-tenant queries possible by design. Isolation lives in the substrate, not in application code or contractual terms.

SOC 2

Roadmap · in discussion

SOC 2 Type II audit roadmap discussion offered with Enterprise design partners. Audit scope available on request.

Region

Australia · Azure

Hosted on Azure Australia regions. Data residency commitments available at Enterprise tier — specify region as part of access conversation.

Graph-native persistent memoryfor production AI agents.