@@ -23,6 +24,8 @@
Multi-Member, Structured Long-Memory, Containerized AI Agent System.
+đ Introduction to Memoh - The Case for an Always-On, Containerized Home Agent
@@ -23,6 +24,8 @@
ć€çšæ·ăç»æćèź°ćżăćźčćšćç AI Agent çł»ç»ă
+đ Introduction to Memoh - The Case for an Always-On, Containerized Home Agent
diff --git a/docs/docs/blogs/2026-02-16.md b/docs/docs/blogs/2026-02-16.md
index 64e32f5b..c29e07f8 100644
--- a/docs/docs/blogs/2026-02-16.md
+++ b/docs/docs/blogs/2026-02-16.md
@@ -6,14 +6,16 @@ author: Team Memoh
# Introduction to Memoh - The Case for an Always-On, Containerized Home Agent
## Overview
-We enter 2026 with a familiar tension: models get smarter every quarter, but the âagent experienceâ still breaks on context, latency, privacy, and real-world workflows. Over the past year, We kept circling three questions:
-- Where does the capability boundary of agents actually sit?
-- Whatâs the real value of long context?
-- What hardware form factor makes âalways-on, personal AIâ feel natural?
+
+We enter 2026 with a familiar tension: models get smarter every quarter, but the âagent experienceâ still breaks on context, latency, privacy, and real-world workflows. Over the past year, we kept circling three questions:
+- Where does the capability boundary of agents actually sit?
+- Whatâs the real value of long context?
+- What hardware form factor makes âalways-on, personal AIâ feel natural?
Memoh is our attempt to turn those questions into something buildableânot a manifesto, but a system that can survive contact with reality.
## Story Time
+
Time travels fast. Somewhere between âIâll remember thisâ and âwait, why did we decide that?â, a year disappears.
Thatâs the annoying part of building: most progress doesnât feel like progress while itâs happening. Itâs just a stream of small choices, half-finished threads, late-night fixes, and the occasional moment that actually clicks. The kind of moment where you sit back and think: okayâthis is real.
@@ -30,30 +32,37 @@ Because the thing LLMs canât give you is not âintelligence.â Itâs weight
Thatâs when I realized what I wanted wasnât âan AI that can talk.â I wanted an AI that can live with youâquietly, continuously, accumulating context without turning your life into content sludge.
-Phones were out first instinctâit's personal, powerful, always there. But mobile OS is closed: without OEM privileges you can build an app, not ambient infrastructure.
+Phones were our first instinctâit's personal, powerful, always there. But mobile OS is closed: without OEM privileges you can build an app, not ambient infrastructure.
-So We looked for the always-on node every home already has: the router (conceptually). Then the economics clashârouter-class hardware canât carry memory, RAG, tools, and multi-user agents. The device evolves: more RAM/storage, a screen, mic/speaker, tiny battery for take out, portable form.
+So we looked for the always-on node every home already has: the router (conceptually). Then the economics clashârouter-class hardware canât carry memory, RAG, tools, and multi-user agents. The device evolves: more RAM/storage, a screen, mic/speaker, tiny battery for take out, portable form.
Eventually it stops being a router. It becomes a new category: a home agent base layer.
## What
+
Memoh is a containerized home/studio AI base layer: cloud-grade model capability paired with local-first memory (knowledge base, RAG/search, conversation history) that stays under your control.
## Why
+
Long-context models raise the ceiling for agentsâbut they also make âfully localâ expensive and âfully cloudâ uncomfortable. People donât want to re-brief AI every day, and they donât want their durable context trapped in someone elseâs feed. Containerization makes Memoh portable, reproducible, and safe to run as always-on infrastructureâso continuity becomes cheap, private, and dependable.
## How
+
We run Memoh as a containerized stack: isolated services for storage (files/DB/vector index), retrieval, tool/runtime execution, and the control plane. Inference calls cloud APIs when you need frontier capability; durable memory and indexing stay local. The device acts as an always-on node (router-like, not a router) serving multiple users with strict boundaries: sharing is explicit, private context remains private, and everything is deployable/upgradable as versioned containers.
## Features
+
- **Multi-bot Management**: Create multiple bots; humans and bots, or bots with each other, can chat privately, in groups, or collaborate.
-
+
+ 
- **Containerized**: Each bot runs in its own isolated container. Bots can freely execute commands, edit files, and access the network within their containersâlike having their own computer.
-
+
+ 
- **Memory Engineering**: Every chat is stored in the database, with the last 24 hours of context loaded by default. Each conversation turn is stored as memory and can be retrieved by bots through semantic search.
-
+
+ 
- **Various Platforms**: Supports Telegram, Lark (Feishu), and more.
- **Simple and Easy to Use**: Configure bots and settings for Provider, Model, Memory, Channel, MCP, and Skills through a graphical interfaceâno coding required to set up your own AI bot.
@@ -61,18 +70,21 @@ We run Memoh as a containerized stack: isolated services for storage (files/DB/v
- More...
## Compare to OpenClaw
-We Shared core belief: both Memoh and OpenClaw treat the agent as more than a chatboxâwe give the LLM a playground: a real environment where it can remember, use tools, and iterate.
+
+We share a core belief: both Memoh and OpenClaw treat the agent as more than a chatboxâwe give the LLM a playground: a real environment where it can remember, use tools, and iterate.
Where Memoh differs:
+
- Lighter and Faster: built as home/studio infrastructure, can be held in the edge device
-- Containerized by default: each bot gets an isolated container (files/commands/network/jobs).
-- Hybrid split: cloud inference, local-first memory + indexing.
-- Multi-user first: explicit sharing and privacy boundaries, support a2a (Agent2Agent).
-- Sustainable: have an experienced team and confidence to push forward and build it.
+- **Containerized by default**: each bot gets an isolated container (files/commands/network/jobs)
+- **Hybrid split**: cloud inference, local-first memory + indexing
+- **Multi-user first**: explicit sharing and privacy boundaries, support a2a (Agent2Agent)
+- **Sustainable**: have an experienced team and confidence to push forward and build it
## Conclusion
+
Memoh is built for one thing: always-on continuityâan AI that stays online, and a memory that stays yours.
We keep frontier inference in the cloud, keep durable context local, and run everything as a containerized, always-on stack. If you want an agent that feels less like an app and more like home infrastructure, thatâs the bet Memoh is making.
-Furthermore, we will continue to operate and permanently open source it, permanently open-source Memoh, making it a product with long impact.
\ No newline at end of file
+Furthermore, we will continue to operate and permanently open-source Memoh, making it a product with long impact.
\ No newline at end of file