SAIHM — Sovereign AI Horizontal Memory

New project. Potentially billions of AI Agents.

“Sovereign AI Horizontal Memory,” a protocol that provides an encrypted, user‑owned, portable memory layer for AI agents. Audience pages target four groups: Individuals: everyday users who want an agent memory they control. Developers: builders integrating memory into agents or apps (open‑source, APIs, protocol docs). Enterprise: compliance‑sensitive organisations (portable, auditable, right‑to‑erasure features). Government: public‑sector deployments (sovereign auditability, no vendor lock‑in). Although consumer and enterprise subscriptions are shipping now, ultimately it is built for AI Agent and other AI memory providers to integrate into their products to solve the context window problem. They can vastly improve context window memory management using the remember, recall, forget, share, and other tools shipped with the MCP server to reduce or eliminate hallucinations and excessive token usage. In-house, our token usage has been reduced by up to 80% using SAIHM and the following prompt: "Maximize SAIHM use to maximize token economy." Globally dispersed swarms of up to 1,000 disparate or homogenous AI Agents can simultaneously share encrypted memories, which has enormous efficiencies and savings potential for many use cases, among many use cases.

Other, SAIHM uses the following technologies and frameworks: Architecture & Protocol - MCP-first design; all tools route through SaihmProtocolRuntime - Sealed-core architecture using COTI V2 Helium garbled circuits Identity - HKDF chain with MPS-PQC-KEY-GEN-v1 → MPS-AGENT-IDENTITY-v1 Data Persistence - GC-5 ShardReadWriteOrchestrator with tiered persistence selected at write time Tools - Eight MCP tools deployed (four memory, two sharing, two governance) - All tools emit GC-14 audit receipts Integration - Supports agents that speak the Model Context Protocol (MCP) SAIHM operates on the public COTI V2 mainnet and is licensed under Apache-2.0.

Whilst SAIHM also pursues adoption by making its product available via its website, the long-term strategy of universal adoption is to gain traction via white-label to AI Agent and other AI memory providers via standards. The following URLs are just a few examples of SAIHM involvement.: https://datatracker.ietf.org/doc/draft-saihm-memory-protocol/ https://saihm.coti.global/standards/nist-ai-rmf-crosswalk/ https://saihm.coti.global/standards/iso-iec-42001-crosswalk/ https://saihm.coti.global/standards/iso-iec-27001-crosswalk/