Open source

The platform is public. The engineering is the value.

We open source what we build. Three projects under github.com/Simple4uhq — the SAE4U Agent (multi-tenant agent platform we run for clients), sae4u-memory (full persistent memory architecture for Claude — sister project), and SHAITS (the self-hosted trust standard we score against). MIT and Apache 2. Free to self-host. Engagement is for the engineering capability that runs them.

Public repositories

Three projects, one philosophy.

Each one solves a real problem we encountered building production AI infrastructure. Each one is something we still use ourselves. Each one is yours to fork.

SAE4U Agent · Apache 2

Multi-tenant AI agent platform

The orchestrator + per-tenant runtime template we built and use for paid AI Infrastructure engagements.

FastAPI control plane on dedicated ops-master, per-tenant DigitalOcean droplets via DO API + cloud-init, Fernet-encrypted bot pool, Telegram-first agent runtime with FTS5 + sqlite-vec hybrid knowledge base. Includes ARCHITECTURE.md narrative documenting what we built, what we learned, and why we walked away from selling it as SaaS.

Python · FastAPI · SQLite · Apache 2 · self-hostable
View on GitHub → Hire us to run it →
sae4u-memory · MIT · v0.2.0

Persistent memory architecture for Claude

Not just an MCP server — the full memory pattern: hooks, rules, prompts, templates, and a customizable peer-coder persona.

MCP server with two-corpus recall (SQLite facts + markdown files in your auto-memory dirs). Plus the architecture we run in production: a 10-min UserPromptSubmit hook that forces brief memory reviews, 10 universal feedback rules (no session-state confabulation, mirror-handoff, no-hardcode, post-write review, …), session-open and session-close prompts paired with the rules, a MEMORY.md index format, and Simple — a generic peer-coder persona with persistent memory awareness. sae4u-memory init --code-full wires the whole pattern into Claude Code in one command. Sister project to SAE4U Agent.

Python · MCP · SQLite + FTS5 · hooks · rules · prompts · persona
View on GitHub → Install from PyPI →
SHAITS · Open spec

Self-Hosted AI Trust Standard

Open specification for trustworthy self-hosted AI agents. We score against it ourselves.

A 125-point scoring framework for self-hosted AI: data residency, audit logging, telemetry policy, encryption at rest, key rotation, dependency provenance. Cloud-only products are out of scope by design. We publish our own self-audit using the standard. Other self-hosted projects encouraged to audit + register.

Open spec · markdown · self-audit framework
View on GitHub → Read the standard →
Open source vs retainer

Same code. Two ways to use it.

SAE4U Agent on GitHub is the architecture — with empty slots for the parts that make it actually useful for a specific business. Anyone can fork it and fill it themselves. Or it ships pre-filled with your business and operated by us, included in the retainer at no extra AI line item.

Open source · self-host

Architecture with empty slots

The orchestrator, the runtime template, the knowledge base schema, the OAuth wiring, the prompt scaffolding — all there. The slots for what your business actually is — your tools, your team's voice, your workflows, your data — come empty. You fill them. You host them. You maintain them.

  • Apache 2 license — fork freely
  • You provision your own infrastructure
  • You wire your own OAuth integrations
  • You write your own SOUL and workflows
  • You pay only your hosting and AI-provider bills
Fork on GitHub →
Retainer · we run it

Filled, deployed, operated

Inside an embedded retainer, the same architecture ships pre-filled with your business. Your tools wired, your voice in the SOUL, your workflows automated, your data flowing. You get a Telegram chat with an agent that already knows your operation — we keep the engineering behind it sharp every week.

  • Included in the $10–15K/mo retainer — no separate AI line
  • We provision and run the infrastructure
  • We wire OAuth and connect every tool you use
  • We tune the SOUL, ship workflows, monitor health
  • You use the agent to track what and how
Discuss an engagement →
Why open source

No proprietary moat. The engineering is the moat.

Most AI consulting firms keep their tooling proprietary. We took the opposite stance. The reason: clients shouldn't be locked into our tools — they should pick us because the engineering is real, not because we hold their AI hostage. When you engage us, you get a deployment of the same code we publish on GitHub. You can fork, audit, run elsewhere if our engagement ever ends. Trust comes from full visibility, not from contractual lock-in.

Practically: this also doubles as our most credible lead-gen channel. Developers find the code, read the engineering, decide we're real, contact us. That's better signal than any landing page can produce alone.

Want to engage us, or contribute, or just self-host?

All three are welcome. Engagement = paid retainer for the engineering team that runs this code. Contribute = PRs welcome on all three repos. Self-host = fork, run, never talk to us.

Book 15 min →