Docs

The practical guide to using and trusting the persistent brain.

Use this page to understand what HOM Local actually stores, how recall is grounded, and when local-first is the right move.

Start Here

HOM Local gives AI work an embedded persistent brain, not just a history view.

It keeps decisions, evidence, sessions, tools, and model choices connected so work can continue without constant manual re-briefing.

Remember

Keep the decisions, constraints, session notes, and evidence that reduce restart time on the next task.

Recall

Get context back with confidence and evidence, not raw text chunks that still leave the operator guessing.

Continue

Resume work across sessions, models, and agents without rebuilding the project context from scratch.

The product promise

HOM helps AI keep the context that makes work move while keeping the persistent brain under the user's control.

Desktop App

HOM Local ships as a desktop app by design.

The goal is a native local workspace, not a web UI. The app is where you chat, inspect memory, manage providers, review agents, and control privacy.

1

Chat is the primary surface

You work with AI directly. Memory, evidence, and confidence appear when they matter.

2

Cortex is the memory surface

Inspect memories, evidence paths, related context, and the way the brain is organizing work.

3

Sessions stay recoverable

Past sessions can be summarized, searched, inspected, and brought back into current work.

4

Settings expose control

Provider choice, local ownership, learning controls, export, and privacy belong in the app, not hidden behind an account portal.

Cognitive Memory

Memory that supports execution, not passive storage.

HOM screens what enters memory, keeps evidence attached, and returns context with confidence so operators can understand what the AI is relying on.

Quality screening

Low-signal material should not pollute the brain. HOM is built around saving what matters.

Evidence-backed recall

Answers should point back to the memories and signals that supported them.

Visible confidence

When evidence is strong, the product can say so. When evidence is thin, the user should know.

No silent forgetting

Old context should remain reachable. Useful context rises, weak context recedes, but memory is not casually erased.

Learning over time

Patterns, corrections, and repeated recalls help the brain become more useful with continued work.

Inspectable paths

The point is not mystique. The user should be able to inspect why a memory came back.

Agentic Work

Agents need memory that survives the session.

HOM Local is not memory plus chat history. It is a workspace where tools, skills, sessions, and agents can share one remembered working state.

A

Tools

AI work can call tools and use workflows without losing the memory of why a task exists.

B

Skills

Reusable capabilities should be available to the agent while still feeding the same memory system.

C

Agents

Agent contributions should be attributable, inspectable, and constrained by the product's memory rules.

D

Projects and sessions

The project is the continuity anchor. Sessions become part of the project brain instead of disappearing into history.

Models

Your model should remain replaceable.

HOM Local keeps the brain separate from any single provider. Use local models for private work, cloud models when useful, and keep the same persistent working memory across switches.

Local models Ollama LM Studio Claude OpenAI-compatible Gemini Custom providers

Why this matters

Changing models should not mean losing the project, the decisions, the evidence, or the learning history.

Editions

Start local. Add hosted memory only when coordination genuinely needs it.

The practical question is simple: where should the brain live today, and who truly needs access tomorrow?

HOM Local

Free personal edition. The brain lives on your machine. Best for individual AI work and privacy-first building.

HOM Cloud Soon

Coming soon for cross-device access, shared workspaces, hosted memory, and support.

HOM Team Soon

Planned for organizations that need governed memory, auditability, residency, and deeper control.

Privacy

Local ownership is the default.

HOM Local is designed so the personal brain starts on the user's machine. Outbound model calls only belong to providers the user chooses.

1

Memory stays yours

Memory is portable and not training data for someone else's product.

2

Export matters

A memory product must let users leave with their memories intact.

3

Connections are explicit

Cloud providers and connected tools should be user-chosen, visible, and revocable.

Read just enough to start proving the workflow.

The docs follow the same rule as the product: keep the important context close and make the next step clear.

See the product Start with HOM Local