assay · audience · what runs without you, and what never does

Two ways to read this, depending on why you arrived.

Assay by Medici is an agent-operated delivery pipeline with a human-governance layer. Five desks take a brief (a work order) from intake to a verified merge, and a person holds the merge and every risk gate. Under it, a change is verified by a party that did not implement it, and the record of that verification is append-only and tamper-visible. What you want to know about that depends on where you are standing, so the site forks here, once, and the paths diverge at the next click.

You run a fleet of coding agents

You are the operator running several coding agents and your own pull-request queue, the team lead whose team is turning into fleet direction, the platform engineer who will install it, the solo builder with one repository, or the reviewer who has to check someone else's agents without installing anything. These are the working questions: what runs while you are not looking, what can never run without you, which agent harness is proven, what setup costs, and how a one-person shop starts.

Running agent fleets →

You are evaluating Assay for an organisation

You are weighing this for a team or a company: the executive who owns the transformation story, the executive accountable for attesting that reported work happened, or the engineering leader who has to make agent delivery survive an examiner. These are the questions a buying group asks: what it changes about control and attestation, what the claim is, and where the evidence lives.

Evaluating Assay for an org →

Start here: the first four pages

Not the whole site: the first four things to read, in order, for each of the two readers above. Both paths reach the glossary at step two, on purpose, so that no term arrives before there is a one-click way to look it up. Four is the count a first-time reader can hold without re-reading; there is more after it, but this is the start.

If you run the fleet

  1. The methodology

    The one mental model, before any vocabulary: the agent that writes the code never gets to say it works, and the party that verifies is not the party that implemented.

  2. The glossary

    The terms the rest of the site uses. Open it in a second tab; you are meant to glance at it, not read it through.

  3. What it costs to adopt

    The setup cost in full, stated above the fold rather than discovered three pages in.

  4. One operated board

    A live board with its dwell times, including the unflattering ones. The first thing on the path a stranger can check without credentials.

If you are evaluating Assay for an organisation

  1. The claim, in one screen

    The claim and its stated boundary together. You arrived to decide, not to install.

  2. The glossary

    Sooner, for you: register, gate, and brief turn up in the first paragraph you read.

  3. What it covers, and what you still bring

    The boundary, not the feature list: the pipeline runs from intake to verified on main, and release, deploy, and rollback are not in it today. Which parts of getting software built Assay covers, and which parts remain yours.

  4. How it runs, with the numbers that exist

    The measured figures, alongside the refusal to publish a productivity multiplier, on the same page.

Still deciding whether to adopt rather than how: the comparison, why not the CI and code owners you already have, answers the question most readers arrive with.