iterion

Why Iterion?

Iterion is a workflow engine for plugging AI pipelines together — making LLMs talk to each other, automating processes, formalising the methods you already use, and evolving all of that fluidly as the work changes.

Origin

Late 2025 / early 2026, frontier models crossed a threshold: structured pipelines (plan → implement → review → fix) started producing output worth coming back to after lunch. “Automate this” became a viable thought rather than a wishful one. Iterion is the engine we built to take it seriously.

A pattern that’s worked for us (one of many)

Plan — written contract from an LLM: goals, files, constraints.
Implement — a tool-using agent (Claude Code, Codex, …) executes with high autonomy.
/simplify — a clarity pass: dead code out, reuse in.
Review-fix — reviewer critiques, fixer addresses, loop until satisfied. For critical or complex work, raise the bar to consecutive approvals across alternating model families.
Light human finalization — real tests, diagonal read.

The same shape works for fixing existing code, and stretches to multi-hour autonomous sessions that produce something near-end-to-end. It’s one pattern; Iterion runs whichever you arrive at.

What Iterion lets you do

Compose pipelines — chain agents, judges, routers, humans, tools, and compute nodes into one graph in a single .bot document (parallel branches converge via await: wait_all / best_effort on a downstream node — there is no separate join node).
Make LLMs talk to each other — multi-agent, multi-backend, multi-model, inside one workflow. Mixing families is a one-line change.
Automate processes — run on demand, on schedule, in CI, or unattended for hours, with budgets and per-run sandboxes. A 90-minute run that dies at minute 80 resumes from minute 75.
Formalise the method that worked for you — the recipe becomes a versioned, diffable, shareable file.
Evolve fluidly — add a node, swap a backend, fork a variant. The DSL is small enough that fluency takes an afternoon.

Measure with the asymptote

Run the same task ten times against the same workflow. Plot quality. The curve climbs, then stabilises — the asymptote. It tells you whether the pipeline converges, what ceiling it converges to, and how much variance to expect on a single run. iterion bench asymptote produces it for any workflow on any corpus.

The asymptote is detected by the judge — its verdict prompt is the load-bearing piece. Treat every new judge as a multi-draft exercise.

Why a dedicated engine

Shell scripts can chain commands but can’t checkpoint long autonomous runs, sandbox each agent, or produce a replayable log. Python frameworks (LangGraph, CrewAI) fit many teams; Iterion picks differently — a small .bot document anyone can read, diff, and re-run without an interpreter. Two recipe variants run side-by-side without touching code.

Get started: install.md.

This site is open source. Improve this page.