Intellectual Property
When we set out to build the organizational mind, we found a chasm — a set of fundamental problems that no existing architecture solves. Each patent represents a structural innovation born from that work. Together, they form an impenetrable architectural moat around governed organizational intelligence.
Portfolio Overview
The LEAPWare patent portfolio is organized into three tiers, each protecting a different layer of the governed organizational intelligence stack. Together, they make replication architecturally infeasible.
2 patents. Foundation-layer inventions that no competitor can replicate without infringing. Governance-enforced knowledge retrieval and experiential knowledge extraction — the two capabilities that make LEAPCortex architecturally unique.
4 patents. Architecture-layer inventions that create platform-level differentiation. Multi-resolution knowledge, VIRT template pipelines, organizational digital twins, and dual-protocol governance serialization.
4 patents. Intelligence-layer inventions that transform stored knowledge into organizational cognition. Disagreement detection, dynamic context assembly, knowledge health scoring, and citation graph brief generation.
Core IP — The Foundation
These patents protect the foundational innovations that differentiate LEAPCortex from every agent memory framework on the market.
US Provisional Application · Filed March 2026
A system and method for evaluating attribute-based access control policies on every knowledge access request — considering agent identity, data classification, organizational hierarchy, task context, and resolution level — before returning results. Sub-10ms policy evaluation per request with full audit logging of every authorization decision.
Key claims: Real-time Cedar ABAC evaluation at the knowledge retrieval boundary. Attribute-set includes agent charter scope, data sensitivity classification, organizational position, and active task context. Deny decisions logged with equivalent detail to allow decisions.
Prior art gap: Mem0 offers tenant-level isolation only. Letta, Zep, and Hindsight have no access control at the retrieval boundary. No framework evaluates multi-attribute policies per-query, per-agent, per-resolution level.
US Provisional Application · Filed March 2026
A closed-loop pipeline for capturing organizational learning as a typed, governed knowledge object. The system creates a Prediction Record at action time, performs semantic outcome matching from the organizational event stream, applies LLM-powered extraction to derive (prediction, outcome) → learning tuples, and stores results as Experiential Knowledge with full provenance chain.
Key claims: Automated prediction-outcome matching. LLM extraction of learning tuples from matched pairs. Storage as typed knowledge objects with bi-temporal versioning. Closed-loop: experiential knowledge feeds back into future agent context assembly.
Prior art gap: No framework in the market captures organizational learning as a governed, typed, retrievable knowledge object with provenance. The organization gets smarter with every decision — not just busier.
Platform IP — The Architecture
These patents protect the architectural innovations that make LEAPWare a platform, not a product.
US Provisional Application · Filed March 2026
A system that stores every knowledge object at four abstraction levels — one-line status, structured summary, detail, and raw source — and dynamically serves the optimal resolution based on consumer identity, context window capacity, and task requirements.
Key claims: Four-level resolution hierarchy per knowledge object. Dynamic resolution selection based on consumer attributes and token budget. Governance-aware: resolution access may be restricted by Cedar policy.
US Provisional Application · Filed March 2026
A two-layer template system for defining governed AI agent roles. Layer 1: seven mandatory charter sections defining authority, skills, expertise, and composition. Layer 2: MCP-native skill packs consumed by pointer reference. The pipeline compiles charter definitions into Cedar policies at instantiation time.
Key claims: Declarative charter-to-policy compilation. Seven mandatory charter sections. Skill pack composition via MCP pointer reference. Governance is architectural — compiled into enforcement policy, not advisory prompts.
US Provisional Application · Filed March 2026
A system that maintains a living, real-time model of the entire organizational knowledge state — types, dependencies, health metrics, contradiction graph, staleness distribution. A queryable digital twin that agents use for planning and humans use for decision-making.
Key claims: Real-time knowledge state modeling. Dependency graph with contradiction detection. Staleness distribution computation. Queryable by both human and AI consumers.
US Provisional Application · Filed March 2026
A system where Cedar policy constraints travel with every task delegation across both MCP (model-to-knowledge) and A2A (agent-to-agent) protocols. When an agent delegates a task to a specialist, the specialist inherits the policy constraints — governance does not stop at the delegation boundary.
Key claims: Policy constraint serialization across protocol boundaries. Governance inheritance on delegation. Support for both MCP and A2A channels. No policy attenuation across the delegation chain.
Application IP — The Intelligence
These patents protect the intelligence layer — the capabilities that transform stored knowledge into organizational cognition.
US Provisional Application · Filed March 2026
A system that reads published intelligence from multiple AI agents and detects four types of disagreement: factual (different conclusions from same data), priority (same facts, different importance), action (same diagnosis, different prescriptions), and timing (same action, different urgency). Each disagreement is surfaced with full reasoning chains for human resolution.
Key claims: Four-type disagreement taxonomy. Cross-domain synthesis boundary. Reasoning chain preservation for human review. Automated escalation on unresolved disagreement.
US Provisional Application · Filed March 2026
A Context Assembly Service that compiles governance-filtered, right-resolution, synthesis-enriched knowledge packages per-request. Evaluates Cedar policies, selects resolution levels, assembles cross-domain context, respects token budgets, and serves both MCP and A2A channels from the same pipeline.
Key claims: Per-request context assembly with governance filtering. Token budget awareness. Multi-resolution selection per knowledge object. Unified serving pipeline for MCP and A2A protocols.
US Provisional Application · Filed March 2026
An autonomic system that continuously evaluates the health of the organizational knowledge graph — staleness, contradictions, orphaned objects, dependency integrity — and computes a single scalar health metric. The vital sign for the organizational mind.
Key claims: Continuous health computation across multiple dimensions. Single scalar metric with drill-down diagnostics. Autonomic remediation recommendations. No knowledge platform in the market has this.
US Provisional Application · Filed March 2026
A system that generates executive briefings where every claim traces back through a citation graph to the source knowledge objects that support it. Not summaries — provable intelligence. The organizational mind reports what it knows and shows the receipts.
Key claims: Automated brief generation with citation graph. Every claim traceable to source knowledge objects. Configurable audience adaptation (executive, technical, audit). Daily cadence with on-demand generation.
IP Strategy
All 17 applications are filed as provisional utility patents with the United States Patent and Trademark Office (USPTO). Non-provisional applications are being prepared for filing within the 12-month priority period. Patent prosecution is managed under the LEAPWare Patent and Trademark Development Standard — a governance document that defines filing criteria, claim strategy, and portfolio management procedures.
LEAPWare intends to file Patent Cooperation Treaty (PCT) applications for core and platform-tier patents, enabling national phase entry in key jurisdictions including the European Patent Office (EPO), Japan Patent Office (JPO), Korean Intellectual Property Office (KIPO), and IP Australia. Filing decisions are prioritized by market relevance and competitive landscape in each jurisdiction.
LEAPWare maintains trademark applications and common-law rights for: LEAPWare, LEAPCortex, LEAPControl, LEAPHive, LEAPCrew, LEAPForge, LEAPCapital, VIRT, and the LEAP Intelligence Cycle (Listen, Explain, Act, Prove). The LEAP logo and Graphite design system visual identity are protected as trade dress.
In addition to patent protection, LEAPWare protects proprietary methods, algorithms, training data curation processes, and operational know-how through trade secret protections including: employee and contractor confidentiality agreements, access controls on proprietary repositories, and need-to-know information distribution. Not all innovations are best protected by patents — trade secrets complement the patent portfolio for implementation-specific methods.
The LEAPWare patent portfolio is held defensively. Our primary objective is to protect our freedom to operate and to prevent competitors from patenting architectural innovations we pioneered. We do not pursue patent litigation as a revenue strategy. Strategic licensing is available for enterprises and partners who wish to build on LEAPWare's architectural innovations under governed terms.
Enterprise licensing of specific patent claims is available for organizations building governed AI infrastructure. Licensing terms are negotiated individually and include field-of-use restrictions, quality standards, and governance requirements consistent with LEAPWare's values. Inquiries: legal@leapware.ai.
The agent memory market — Letta, Mem0, Zep, Hindsight, LangMem, Cognee — is racing to add features. But features are not defensible. Architecture is. These 17 patents protect structural innovations that make governed organizational intelligence possible — not surface features that any team could replicate, but deep architectural patterns that define how knowledge is stored, accessed, governed, synthesized, and served. The moat is not code. The moat is the inventions the code implements.
For investors: this portfolio represents a defensible technical foundation that compounds in value as the governed AI market matures. For enterprise buyers: these patents mean the architecture you build on is protected — your investment in LEAPWare is not at risk of being commoditized by copycats. For competitors: we welcome innovation in this space, and we will defend our innovations through every available mechanism.
Patent Pending — All applications filed as provisional utility patents with the United States Patent and Trademark Office. Classification, filing strategy, and portfolio management governed by the LEAPWare Patent and Trademark Development Standard. For IP inquiries: legal@leapware.ai