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.
Tirrekp•1mo ago
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.”
Tirrekp•1mo ago
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.