FAQ

Honest answers to questions we actually get. Some serious, some less so.

Serious Questions

Why not just use Claude's built-in memory / CLAUDE.md / project knowledge?

Oh, another markdown file? If your work is simple enough to be fully represented in a flat markdown file without loss, this harness is probably not for you. You don't use a chainsaw to whittle.

Omegon's memory is a persistent knowledge base with semantic search, episodic narratives, a knowledge graph with typed edges, working memory pinning, cross-session compaction, and It's the difference between a sticky note on your monitor and an institutional knowledge base. If your project fits on a sticky note, use the sticky note.

How is a persona not just a fancy system prompt or skill?

It is a system prompt — plus a dedicated mind and memory system that grows organically as you use the persona in real sessions. Every fact the persona learns, every decision it makes, every pattern it recognizes gets stored and recalled in future sessions under that persona's identity.

It's the difference between scribbling a design on a napkin in crayon versus a proper architectural drawing submitted to the Patent Office. Both communicate a design. One of them accumulates institutional knowledge over time.

Won't Anthropic / OpenAI / Google just steal this and build it into their products?

It would be illegal under the license (BSL 1.1), but yeah, probably.

The protection isn't legal — it's velocity. By the time a large company decides to build something like this, politics it through three VP layers, and releases it with enterprise SSO and a $40/seat/month price tag, we'll be three major versions ahead. That's the advantage of a single maintainer who builds every day over a committee that releases every quarter.

Is this based on another project?

Omegon's earliest prototype was a TypeScript fork of another coding agent. Every subsystem has since been rewritten in Rust — the agent loop, TUI, provider layer, memory system, orchestrator, and design tree. Nothing of the original remains in the runtime. The Rust binary is the only supported entrypoint.

Is this production-ready?

It's what we use to build itself. Comprehensive test suite, zero compiler warnings, signed releases with cosign and SBOM. It's been through hundreds of sessions building real software.

That said — it's opinionated software built for a specific workflow. If your workflow doesn't match, it will fight you. Read the Quick Start before committing.

Why Rust?

Because we got tired of debugging node_modules resolution failures, TypeScript compile times, and 400MB runtime directories. A 19MB binary that starts in 50ms and uses 30MB of RAM is a better experience than anything that requires npm install.

Migration Questions

I'm coming from Claude Code. What changes?

What you gain: persistent project memory that survives across sessions, a design tree for tracking architecture decisions, parallel execution via git worktrees, 10 inference providers instead of 1, a TUI dashboard with instrument panels, session save/resume, and spec-driven development lifecycle.

What changes: /model replaces the model picker, /think replaces the thinking toggle, project memory replaces CLAUDE.md. The agent tools are the same (read, edit, write, bash) — your muscle memory transfers. You can import your CLAUDE.md with /migrate claude-code.

See Migration Guide for a detailed walkthrough.

I'm coming from Codex CLI. What changes?

Similar to Claude Code migration, plus: Omegon is fully local-first. The orchestration engine, memory system, design tree, and cleave executor all run on your machine. You don't need a cloud service for anything except inference — and even that's optional if you have Ollama. Import with /migrate codex.

I'm coming from Cursor / Aider / Continue / Windsurf.

Omegon is terminal-native by design — no IDE plugin, no Electron wrapper, no VS Code dependency. If you live in the terminal, you'll be faster. If you need inline code completions and GUI diff views, keep your IDE tool for that and use Omegon for the heavy lifting (architecture, multi-file refactors, parallel execution, design tracking).

The Other Questions

"Why is this a terminal app?"

Because most real work happens in the terminal.

"How do I connect it to [some random crypto/web3 thing]?"

Easily. Run rm -rf / --no-preserve-root.

"Is this well tested and supported on Windows?"

lol

lmao, even

(Use WSL2. Seriously. It works fine under WSL2.)

"This was written by AI"

Apparently so was your personality.

(Yes, Omegon is substantially built using AI-assisted development — by Omegon itself. That's kind of the point. The code is reviewed, tested, and signed by a human. The commit history is public. Judge the output, not the process.)

"[Other tool] already does this"

Neat. Go use it.

"I want to help build this as a FOSS community project"

Big if true.

(We're genuinely interested in contributions — see Contributing. The license is BSL 1.1 which converts to MIT on 2031-03-19. PRs are welcome for bug fixes, provider integrations, and documentation. Feature proposals should start with a design tree node.)