What using Ecko actually looks like

Day one — about fifteen minutes

One person with admin rights does the setup. Everyone else is unaffected and does not need an account yet.

• Connect GitHub. Ecko installs as a GitHub App and reads PR comments, review comments, and closed issues on the repositories you select. You choose which — it is not all-or-nothing.
• Connect Slack. Pick the channels worth reading. Most teams start with two or three where architectural argument actually happens, not every channel.
• Invite the team. Anyone who should read decisions needs an account; they do not need to do anything else.

There is no schema to design, no template to agree on, and no backfill step. Ecko starts from the next conversation your team has.

What changes for everyone else: nothing

This is the part worth being clear about, because it is the difference between a tool that gets adopted and one that gets abandoned in month two. Your engineers do not write records. They do not tag messages, fill in a template, or remember a process. They argue in pull requests and Slack exactly as they already do.

Ecko reads those threads on a schedule and extracts what was decided, why, and what was rejected. The record exists because the discussion happened.

Who does what

• Admin — connects integrations, manages billing, invites people, and deletes the organization. Usually one or two people, often a tech lead or eng manager.
• Member — reads and searches everything, records decisions by hand when one happens somewhere Ecko cannot see (a meeting, a whiteboard), and owns assumptions on decisions they made.
• Viewer — read and search only. Useful for people who need the context but should not change the record: product, design, incoming contractors, a new hire in week one.

The one habit worth forming

Ecko extracts assumptions automatically, but a human is better at knowing which ones are load-bearing. The teams that get the most out of it do one small thing: when an alert says an assumption looks stale, someone confirms or dismisses it rather than letting it sit.

That is roughly a few minutes a week, and it is what keeps the system an early-warning mechanism rather than another archive nobody trusts.

What the first month looks like

• Week one — decisions start appearing. Expect to disagree with a few extractions; correcting them takes seconds and is the fastest way to see how it reads your team's writing.
• Week two to three — search becomes the point. The moment it earns its place is the first time someone asks "why is it like this?" in a channel and gets a link back within a minute.
• Week four — the weekly digest starts being read rather than skimmed, because by then it is summarising decisions people remember making.
• Month two onwards — the assumption alerts begin firing on things decided in week one. This is the part no wiki does.

What it does not do

Worth stating plainly so nobody is surprised. Ecko does not make decisions, approve them, or enforce a process. It does not read private DMs. It does not import your existing wiki — it starts from the point you connect it, and the archive builds forward rather than backward. And extraction is a language model reading prose, so it is very good but not perfect; the record is editable for exactly that reason.