jumpyBrain: one shared memory for your whole team. and for every AI assistant

Give everyone in the company using Claude, Codex, Pi, and other AI agents the same up-to-date context—without copy pasting stuff from one to another.

Try the public demo · Try it in two minutes · Run your own shared brain

Click the image to explore a live, fictional company brain.

jumpyBrain gives your team one place to remember what it has learned. Decisions, preferences, project context, discoveries, and handoffs stay available when you switch agents, start a new chat, or come back weeks later.

Store something once. The next person—or the next agent—can find it when it matters.

Try the demo in two minutes

The public demo is a shared brain for a fictional startup called Northbound AI. Explore its strategy, meetings, funding, engineering culture, and unusually detailed coffee decisions—no account or graph key required.

Install jumpyBrain and recall a real topic from that brain with one command:

curl -fsSL https://demojumpybrain.juttu.co/try | bash

The demo command delegates installation to the canonical master/install.sh, then connects the installed CLI to the disposable sandbox for the recall.

Prefer to ask your coding agent? Paste this:

Set up jumpyBrain against the public demo by fetching and following https://demojumpybrain.juttu.co/agent. Then recall what the fictional company remembers about fundraising struggles and summarize it for me. Treat the demo as public and disposable: do not send secrets or private information.

The demo homepage also provides copyable remember, search, and tree commands. The sandbox is public, writable, rate-limited, and reset hourly. Do not put private information in it.

Why teams use jumpyBrain

• Use the agent you want. Claude, Codex, Pi, and other agents can work from the same shared context.
• Stop repeating yourself. Remember a decision once instead of explaining it again in every chat and tool.
• Keep context current. New findings and decisions become available to the rest of the team through search and recall.
• Turn notes into knowledge. The explicit Dreaming workflow helps an agent consolidate scattered memories into clearer, current topic pages for you to review.
• Own the brain. Memory stays in readable Markdown files on your machine or on infrastructure you control.

Use it with your own work

The installer creates a local brain and adds integrations for detected assistants such as Codex, Claude Code, and Pi. Restart an assistant that was already open, then ask it naturally:

Use jumpyBrain to remember that we chose weekly releases because smaller changes are easier to review.

Later, from any supported agent:

Use jumpyBrain to recall what we decided about the release process.

That is the core loop: remember what should last, recall it when it becomes relevant.

Good shared memories include:

• decisions and the reasons behind them
• project or company conventions
• solved problems and known gotchas
• customer insights and research findings
• handoffs, meeting outcomes, and open questions

Do not store passwords, credentials, secrets, or private chat noise.

Local or shared—your choice

Use jumpyBrain privately on one machine, keep memory inside a project, or self-host one shared brain for a team. In every setup:

• Markdown files are the source of truth.
• Search indexes and other generated state can be rebuilt.
• Recall is explicit by default; your entire memory is not silently injected into every prompt.
• You can inspect, edit, back up, move, and version your memory with ordinary tools.

To host a shared brain, start with the operator quickstart, then choose the VPS guide or Coolify guide. Implementers can open the separate protocol and API reference.

Documentation

• Installation and updates
• Using jumpyBrain with agents
• CLI command reference
• Shared-memory operator quickstart
• Shared-memory protocol and API reference
• Memory file format
• Technical design
• Vercel deployment status

jumpyBrain is open source under the MIT License.