ortim.dev
Documentation menu

Ortim CLI Documentation

Ortim is auditable AI coding — built for software agencies that answer to clients and teams that answer to regulators. Every run produces reviewed code and a tamper-evident audit trail: who approved what, why, and what the model actually did. Under the hood, a disciplined multi-agent software factory turns a one-paragraph brief into working code — without surrendering control to the LLM.

pip install ortim
ortim demo        # full pipeline walkthrough — no API key required

Requires Python ≥ 3.11. The demo replays a recorded real run when no key is configured, so you can see the whole flow at $0 before any setup. For real projects, ortim config init picks a provider (DeepSeek / Anthropic / local Ollama) and stores credentials — no .env setup needed.

What it is

Ortim treats AI coding as a production-engineering problem, not a prompting problem. Five structural pieces turn an unreliable LLM into a reviewable software pipeline:

  • A deterministic state machine drives the pipeline: brief → PRD → RFC → task DAG → code.
  • Two mandatory human gates (PRD and RFC) — the LLM never silently commits to a scope or an architecture.
  • A hash-chained audit log records every LLM call, state transition, and gate decision.
  • Scope-locked task DAGs keep work inside declared module boundaries.
  • A reviewer chain (Code → Security → Test → Perf) that treats "couldn't verify" differently from "passed".

Where to go next

  • Getting started — install to a finished project in ~15 minutes.
  • Concepts — the state machine, gates, deterministic architecture, and audit chain.
  • CLI reference — every command you'll use day to day.
  • Providers & cost — multi-provider routing and what a run actually costs.
  • Ortim Cloud — sync runs to a team workspace and share read-only audit links with clients.

What it is not

Ortim is not an autocomplete or an IDE. If you love Cursor or Claude Code for interactive work, Ortim is the opposite end of the spectrum: batch, gated, auditable. It works on greenfield and existing codebases — brownfield mode scans the import graph. The bet is that teams who need to answer "who approved this, and why?" will trade some interactivity for governance.