- Research
- API & Docs
- Mind Platform
-
Our Mission
- Support
Mpalo builds memory systems that grow with use. Privacy-first storage that evolves through every interaction, giving your models context that persists and adapts. Memory designed for continuity. Built to remember what matters.
An open-source agentic coding runtime. Research-driven, terminal-first, built for the developer who wants to see every decision the system makes.
PaloSpring is currently in beta.
Most memory systems extract facts from conversations and retrieve the most similar ones next time. Palo encodes experience, not just what was said, but when, in what sequence, in what context. The difference is structural: Palo preserves continuity across experiences instead of reducing a person to a list of extracted facts.
Not everything deserves equal weight. Palo distinguishes what matters from what doesn't, holding what shapes a person's context over time, releasing what doesn't. Memory that forgets intelligently stays useful. Memory that keeps everything becomes noise.
When something happened matters as much as what happened. Palo treats temporal position as a first-class property of every memory, not a timestamp attached to a fact, but a structural dimension of the experience itself. This is what makes chronological reasoning possible.
Palo learns the patterns that recur: how a particular user approaches particular problems, what their workflows look like, what they return to. And it recognises when something genuinely new has arrived, updating rather than overwriting.
The what, the when, and the why are encoded together as experience rather than extracted as separate facts. Episodic structure lets an agent maintain continuity across a user's history instead of treating each interaction as isolated.
Palo is priced by Memory Operations. Palo Nano and Palo Bloom are designed for affordable production memory, while Palo Research is priced for long-horizon synthesis and depth. External LLM usage remains separate and is billed through your own provider keys.
Your data stays yours. Connect your own vector store or use Private Memory Spaces for isolated, user-scoped storage. Palo builds memories from your data. Your raw content and memories remain under your control and are never used to train shared models without your explicit consent.
