YAGNI Code · internal preview

A terminal coding agent that already knows your company.

Like Claude Code or Codex, but wired into YAGNI. It starts every task grounded in how your company actually works, so you step in less.

One YAGNI login. The model is routed for you, you bring no API keys.

Same workspace your Teams, Briefs, and Feed already run on.

Book a demo Start with the business workspace see how a run works

Boot grounding + ask_yagni

Read the room before the first line.

A blank agent starts from amnesia. YAGNI Code boots on a brief of your company, its vision, goals, conventions, and the why, and works from it right away.

On startup it loads how your company actually works. No prompting, no pasted context.

Mid-task, the ask_yagni tool pulls a cited answer from the YAGNI app instead of guessing on anything specific to your company or codebase.

The bet: an agent that consults your conventions, decisions, ownership, and priorities ships the right thing more often than one that improvises.

A blank agent

Payments service, I think. Safest to just re-run the charge.

no source
ask_yagni YAGNI app
agent asked

which Team owns billing retries, and what is the convention for partial charges?

cited answer

The Billing Team owns retries1. Partial charges follow no double-charge on retry: reconcile to the amount already captured, never re-authorize the card2.

1Billing Team · Brief 2decision: no double-charge on retry

/go

From a ticket to a reviewed change.

Point /go at a ticket and walk away. It runs a grounded loop, map, plan, implement, review, fix, and hands back a change you can read.

A live activity feed shows the stage it is on while you do something else.

Every run is recorded on the Work page, so there is a trail of what it did and why.

Run YAG-318 · live auto
Billing Team · reconcile credit-ledger drift
map 12 files mapped
plan plan: 4 steps
implement 7 files changed
review reviewing diff
fix queued
Recorded on the Work page when it finishes.

review_business_match + record_engineering_context

It reviews for the right thing, then writes back what it learned.

Most tools check for bugs. YAGNI Code also checks the diff against your business. It flags correct code that does the wrong thing and cites the Team decision it conflicts with.

Correct code, wrong thing. It names the decision the change breaks instead of waving it through.

When the work lands, it records the engineering why back onto the originating work item. Real business signal in, real consequence out.

review · /go run · billing-service
Business match: this looks off.
services/billing/retryCharge.ts
async function retryCharge(invoice)
+ await charge(invoice.card, invoice.amount)

The retry path charges the card a second time. The code is correct, but it conflicts with a standing decision.

Billing Team decision: no double-charge on retry.

cites · Playbook
Recorded on YAG-318

Why Kept the idempotency key on retry to honor the no-double-charge rule.

Captured corrections

Every correction makes it sharper.

When a human edits the agent's PR before merge, those edits are captured as labeled corrections. They ground the next run, so the same mistake does not come back.

It gets better the longer your team uses it. Your edits are the training signal.

This is the part a fresh install cannot copy. The moat is your accumulated corrections.

Engineering Team · captured from your edits
Pre-merge edits per run

Your PRs needed 5 edits a run in May. This week, 1.

Fewer human edits over time. That is the bet, made visible.
Learned
Captured from your pre-merge edits · 3 corrections

prefer the resilientFetch seam over raw fetch

Convention · from PR #1227

name the Team decision in review notes

Review · from PR #1231

no double-charge on retry, always

Billing · from PR #1228
Each correction grounds the next run.

/mode · /cost

You stay in control. It earns more rope.

Autonomy is earned, not assumed. Each agent climbs a ladder, Training, Supervised, Autonomous, and you decide when it moves up.

Set the permission mode with /mode: plan, review, or auto. This run is in review, so it proposes and waits for you.

Set a ceiling and watch spend with /cost.

Nothing irreversible ships without you. Routine autonomy, permanent human floor.

Payments Engineering · earned autonomy
Autonomous not yet reached
Supervised current
Training
/cost
spend $4.10 of $20.00 ceiling within ceiling

One workspace

Part of your autonomous business.

YAGNI Code is not a separate tool. It runs in the same workspace as your Teams, Briefs, and Feed. One login routes the model and the business context.

Runs show up on the Work page next to everything else the company is doing.

The model is invisible, you bring no API keys, and it is priced per workspace, not per seat.

It is the engineering organ of an autonomous business.

Book a demo
Work
Engineering Team
YAG-318 retry idempotency in review
YAG-301 webhook backfill merged
YAG-296 cost ceiling copy merged
Model Routed for you
Pricing Per workspace

One shared record. Not a bolt-on.

Internal preview

The engineering organ of an autonomous business.

When the work is code, YAGNI Code ships the code. Grounded in your company, reviewed for the right thing, recorded on the Work page.

Open access is by demo today. Bring a ticket. Watch it come back reviewed.

Book a demo Start with the business workspace

A terminal coding agent that already knows your company.