# Find the memory.

Korium is shared company memory for people and AI agents. It keeps your team’s decisions with their reasons and sources, so connected assistants can find what changed, what failed and what worked before starting the next job.

[Sign up for Korium](https://app.kyroco.ai/pricing.md) [Already have an account? Set it up](docs/korium/connect-korium.html)

Remember / Locate / Equip / Run

## Memory is where it starts. The work is where it matters.

Models bring intelligence. Tools bring capabilities. Korium brings the memory, the code map, the ready-made skills and the workflows that put it to work. We call it Collective Intelligence.

1. 01 / Remember
   
   **Korium:mem**
   
   ### Shared memory
   
   What you know, tried and learned.
2. 02 / Locate
   
   **Korium:code**
   
   ### Code intelligence
   
   Find the right code, at the right commit.
3. 03 / Equip
   
   **Korium:skills**
   
   ### Ready-made skills
   
   Load the capability the work needs.
4. 04 / Run
   
   **Korium:workflows**
   
   ### Reusable workflows
   
   Use them, change them or add your own.

Work completes. Learning returns to memory.

Use what’s there, change it for your work or build your own. Save back what you learn so the next run starts further ahead.

[How the four parts work together](docs/korium/capabilities.html)

Keep more than the answer

## You need to know why.

“We chose this approach” is a start. “We chose this approach because the other one failed under load, and here’s the test” is something the next person can work with.

### Keep the decision and its source.

Save what was decided, who made the call, the evidence and which project it belongs to. An agent’s summary and a decision approved by a person are different things. The record should say which is which.

### Keep what changed.

When a decision is replaced, connect the new memory to the old one. You can ask what’s current or follow how the decision changed. No one has to pretend the earlier work never happened.

### Find the relevant part.

Search combines the words you used with what you meant. Importance and use help order the results. You get the memories behind the answer so you can check them.

### Use your own language.

Your team already has words for its work. Use those names for projects, responsibilities and relationships. A memory system should understand how you talk about your company.

[Write, search and correct memories](docs/korium/working-with-memory.html)

One memory across connected tools

## Change the tool. Keep what you learned.

Claude / OpenAI / Gemini keep memory inside their own platforms. By design, saving something there does not make it shared company memory. Korium gives connected assistants a place to read and write the same learning.

**Assistants** Claude / OpenAI / Gemini / OpenClaw OpenClaw tested

**Transcripts** Teams / Zoom / Meet

**Korium** Company memory

**Work tools** Miro / Atlassian / Figma

**Agents / workflows** Read what matters. Save what changes.

Sources reach Korium through a connected assistant, an import or an integration. This diagram does not mean every tool has a built-in connector.

Meeting notes, transcripts, documents and work from other tools can become memory when you bring them into Korium. You decide what to save and who can use it. Connecting Korium does not give it blanket access to every service your company uses.

[Connect an assistant](docs/korium/connect-any-mcp-client.html)

For teams building software

## Your agent shouldn’t have to rediscover the whole codebase.

Korium:code gives it a map of the code: the symbols, the files and the line ranges, tied to a commit. Korium memory gives it the decisions and lessons that explain why the code works that way.

### The source stays on your Mac.

The Mac CLI extracts symbol names, locations and fingerprints. It sends that map to Korium, not the source text. The agent still reads the relevant code locally before changing it.

### Freshness is part of the answer.

A result names the commit it came from. If that differs from your checkout, verify the location. Use a Git hook or a task-end hook to keep the map up to date.

[Index your first repository](docs/korium/indexing-your-code.html)

A habit your agents can follow

## Read before deciding. Save before handing off.

1. Task
2. Read memory
3. Decide
4. Act
5. Store decision
6. Hand off

The read and the write go to the same company memory.

Give your agent the instructions, skills and hooks to follow this pattern. When a connection fails, it should tell you. When a save fails, it should not claim the learning was kept.

[Build memory into your workflow](docs/korium/make-it-automatic.html)

It’s your company’s knowledge

## Treat it that way.

Korium separates tenant data and controls access on the server. The connected client’s permissions determine which tools it can use. You can give an assistant read access without granting it permission to capture new memories.

Keep credentials and private information you don’t need out of memory. Review sources before treating an extracted statement as a company decision. Our published security and privacy documents explain the data handling and the terms that apply.

[Access and data handling](docs/korium/security-and-access.html) [Security overview](legal/security.html)

Start simple. Go deeper when you need to.

## Instructions you can actually follow.

1. 01 / Getting started
   
   ### Save one memory. Find it again.
   
   Connect your assistant, save a useful decision and retrieve it from a fresh chat.
   
   [Start here →](docs/korium/getting-started.html)
2. 02 / Using Korium well
   
   ### Give the next person better context.
   
   Write useful memories, keep their sources, handle corrections and put skills to work.
   
   [Use Korium well →](docs/korium/index.html#advanced)
3. 03 / Technical guides
   
   ### Connect it to the way you build.
   
   Codex / Claude Code / Bash / Git / GitHub. Working examples and the checks that matter.
   
   [Read the technical guides →](docs/korium/index.html#technical)

Put it to work

## Start with something you don’t want to explain twice.

Connect your assistant and save a real decision. Then start a new session and ask for it back.

[View plans & sign up](https://app.kyroco.ai/pricing.md) [Manage subscription →](https://app.kyroco.ai/account) [Talk to Robert ↗](https://fantastical.app/kyroco/meet-with-robert-sfeir)

Compare subscription plans without signing in. Choose your plan, then create your account.
