← midhunkumar05.github.io checking version…

Self-hosted · Enterprise scale

Eunomia

A Super Intelligence Agent AI memory system, built for enterprise scale in Rust — vaults, an entity/memory graph, connectors, and a tool-calling chat agent that reasons over what it's accumulated instead of starting from zero.

Install

Works on macOS and Linux (Windows: inside WSL). Anything missing, Docker included, gets installed for you. Takes a few minutes.

curl -fsSL https://midhunkumar05.github.io/eunomia/install.sh | bash
Paste into a terminal. It asks a few questions, and Enter accepts each default.
Customise the command

What the installer does

  1. Checks your machine and installs anything missing: git, Docker (Colima on macOS, so no GUI), and jq. It starts Docker if it isn't running.
  2. Asks a few questions: folder, ports, an optional model endpoint (base URL first, then key), updates, your account, and optionally HTTPS at your own domain with a free Let's Encrypt certificate (--domain, --acme-email; or later in Settings → HTTPS). Then it shows a summary before changing anything.
  3. Installs: clones the latest release, generates secrets into .env, and starts SurrealDB, the API and the web app.
  4. Connects: turns on one-click updates (Settings → Updates) and adds Eunomia's MCP server to Claude Code, Codex, Hermes, Gemini CLI, Cursor, Windsurf, OpenCode, VS Code and Claude Desktop, each with its own token. It also installs the memory skill, and for Claude Code adds automatic recall on every prompt.

A model key (OpenAI or any compatible endpoint) is optional. It's recommended to save your agent's tokens: Eunomia then does embeddings and answer synthesis itself.

Docs

What it does

Eunomia is a Super Intelligence Agent AI memory system, built for enterprise scale in Rust. At its core is an entity/memory graph: people, organisations, locations, and (for working in codebases) repositories, files, and symbols — each with notes attached and relations drawn between them. A vault system scopes that graph per-user or per-team, so a personal knowledge base and a shared one coexist without stepping on each other.

A chat agent sits on top with tool-calling access to that graph plus a recall/search layer over ingested records, so it can look things up, write new memories, and cite its sources rather than hallucinate them. Connectors (Up Bank, PocketAI, and a generic open connector) pull external data in on a schedule, and every mutating action is written to an audit log.

A memory graph, in practice — entities synced from your data, linked by typed relations:

influenced worked_at located_in Ada Lovelace PERSON Alan Turing PERSON Bletchley Park ORGANISATION London LOCATION
Person Organisation Location

The same graph covers code — repositories, files, and symbols, linked the same way:

contains defines calls Eunomia REPOSITORY auth.rs FILE login() SYMBOL verify_token() SYMBOL
Repository File Symbol

Stack

Backend

Rust, Axum, the official SurrealDB SDK — auth, vaults, entities, connectors, chat streaming (SSE), all on one schema.

Database

SurrealDB — graph relations, full-text search, and vector search (MTREE) in one engine.

Frontend

Next.js, TypeScript — dashboard, entity graph, chat, vaults, connectors, settings.

Deployment

Docker Compose — SurrealDB, backend, and frontend, one command to bring up.

Updates

Every install can update itself: Settings → Updates shows when a new release is out, and Update now applies it within a minute (prebuilt images, no rebuild). Scripts can check the latest release too:

GET https://api.github.com/repos/Qyrhal/Eunomia/releases/latest

Source

Code is on GitHub.

github.com/Qyrhal/Eunomia ↗