Quick start

Get eden-memory running locally and confirm that your agent can remember and recall a fact across sessions.

Prerequisites

A Linux or macOS machine.
Shell access and curl.
An MCP-compatible agent or editor.

1. Install

Run the installer:

curl -fsSL https://0d3sa.com/eden-memory/install.sh | sh

This downloads the right binary for your platform, verifies its checksum, and installs it to ~/.local/bin/eden-memory. If your terminal is interactive, the installer will prompt you for an EDEN_ORG_ID and write it to ~/.eden-memory/.env. You can leave it empty and configure it later.

For non-interactive installs, set the organization ID ahead of time:

export EDEN_ORG_ID=your-org
curl -fsSL https://0d3sa.com/eden-memory/install.sh | sh

If you prefer a manual install, see Downloads and checksums.

2. Verify

Check the binary:

eden-memory version

Then confirm it can open its database:

eden-memory health

You should see a version string and a health report with status: ok. If either command fails, make sure ~/.local/bin is on your PATH, or use the full binary path.

3. Connect your agent

eden-memory speaks MCP over stdio. Pick the tutorial for your client:

If you already know your client’s mcpServers JSON, the server command is:

/home/yourname/.local/bin/eden-memory --db /home/yourname/.eden-memory/default.db

Replace /home/yourname with your actual home path and use absolute paths. Then restart your client.

4. Remember and recall

Ask your agent to remember something:

Remember that I prefer Python examples and short sentences.

Then start a new conversation and ask:

What do you know about my communication preferences?

The agent should recall the preference from the local store.

Expected output

eden-memory version prints a version string.
eden-memory health returns status: ok.
A remembered fact is returned when asked in a new session.

What eden-memory does

eden-memory stores memories in a SQLite database at ~/.eden-memory/default.db. Each memory gets a 256-dimensional embedding. When the agent runs eden_recall, eden-memory compares the query embedding to stored vectors and returns the closest matches.

[!TIP] Short answer eden-memory is the required base for every user. If you work in Claude Code and want structured role-based collaboration, add the Agentic Team Protocol on top.