Tirrekp

We’re building a Memory OS — infrastructure that gives AI agents persistent state, truth guarantees, and long-term continuity across interactions.

Looking for:

• distributed systems engineers • infra engineers comfortable with logs, state stores, and latency optimization • people who like building primitives and invariants, not features

We’re in pilot with a logistics-focused design partner and expanding.

Interested? DM me.

Hi HN, I am a 20 year old founder from Louisiana working on MotionOS.

MotionOS is a shared memory layer that sits under AI contact centers and voice agents. Right now most stacks treat each call as a fresh start. History is scattered across tickets and transcripts, so bots and agents keep asking customers to repeat themselves.

MotionOS keeps a single timeline per customer, extracts key memories and decisions from transcripts, and exposes simple APIs so agents and bots can: • recall the story of a caller at the start of a conversation • update that memory at the end of each interaction • track continuity metrics like repeat contact rate

I am very early. The site shows the concept and the first pieces of the system I am building. I would love feedback on: • whether this memory layer seems useful in real contact center stacks • what data models or integrations you would expect • gotchas you see from your own experience with support tools or AI agents

Thank you for reading and for any blunt feedback.

We’re building a shared memory layer that sits between AI agents and CCaaS tools Zendesk, Dialpad, Five9, so every call and ticket shares the same history and “story of the customer.”

I’m a 20-year-old founder building MotionOS – a shared memory and state layer for AI agents, with our first real focus on voice AI and call centers.

Right now most AI call / support systems still treat every conversation like a fresh session. Context is spread across chat history, tickets, CRMs, and ad-hoc notes. When the bot hands off to a human (or the customer calls back next week) a lot of what happened is lost.

MotionOS tries to make “memory” a first-class infra layer instead of an afterthought: • per-caller timelines that merge events from calls, tickets, and tools • APIs to read/write memory in a structured way from any agent stack • policies for summarising, forgetting and governing long-term context • a simple dashboard to inspect what an agent “remembers” and why

The current site still shows the more general “developer” version of MotionOS, but our design partners are all in voice AI / contact center land, and we’re shaping the product around those workloads first.

What’s live today: • basic REST/SDK APIs for logging events and querying timelines • a minimal dashboard to explore memory for a given caller / entity • early integrations for typical AI agent stacks (Retell/Vapi-style flows)

What I’m looking for from HN: • feedback on the data model (what’s missing for real-world agents) • critiques of the API design and dashboard UX • pointers from people running production AI voice / CCaaS systems

If you’re building agents (especially for calls / support) and the idea of a shared memory layer seems useful or misguided, I’d love to hear your take.