Did It Though?Sign inStart free

Visual proof for agent-built software

Your agent says it nailed it. Did it though?

Every change your agent makes arrives with a screenshot or a screen recording attached. Point at what is wrong and say why. The note goes back with the coordinates on it — so the agent fixes what you pointed at, not what it guessed.

Start free

Free forever plan · no card · works with any agent that can call an HTTP API

See how it works

How Did It Though? works

01 — Ask

You ask for something big, and leave it to it.

A whole checkout flow. You go and do something else while it works.

Features and benefits

Everything the loop needs, and nothing that gets in its way.

123

Annotate

Point at it. Don’t describe it.

Drop a pin, drag a rectangle or throw an arrow at the thing that is wrong, and say why. Zoom to full resolution first if the problem is three pixels wide. Recordings take notes the same way: say “here, the sheet jumps” at 7.4 seconds and the moment is the comment.

Preview

See it before you run it.

Screens and walkthroughs are in the review at full resolution, the moment your agent finishes. No checking out the branch, no spinning up a preview environment, no waiting on a deploy to find out whether the thing even looks right.

{
  "kind": "rect",
  "geometry": {
    "x": 0.07, "y": 0.62,
    "w": 0.86, "h": 0.08
  },
  "revision": 1,
  "status": "open"
}

Handover

Your agent reads it, not you.

Threads come back as JSON with normalised 0–1 geometry against the revision you drew on, so your agent knows which pixels you meant — without you screenshotting anything, pulling the branch, or running it yourself to explain what is wrong.

pixelhop/chatly #128[![checkout-390.png](https://diditthough.app/e/…)](…)
Verified

Pull requests

Evidence in the PR. Nothing in the repo.

Every upload returns an embeddable snippet. Drop it in the PR body and the images and recordings render where the review already happens — while the files live in object storage, so your git history stays code.

Sharing

One link, anyone you like.

Send a read-only URL to a teammate or a client and they are in — no account, no seat, no invite. It is a capability link, so it is never indexed, revocable in a click, and it never exposes the underlying file.

Pricing

Two plans. Priced on evidence, not on seats you forgot about.

You are charged for the evidence we store and serve, not for the people looking at it. A reviewer joins on a link with no account and no seat; an image is one evidence unit, a video counts by its length, and re-uploading the identical file counts for nothing.

Free

For one person putting one project under the microscope.

$0forever

Start free
Evidence units
500 / month
Reviews
25 / month
Named members
1
Media retention
14 days
  • 3 agent keys across 3 projects
  • Screenshots and video up to 5 min
  • Structured feedback over API, CLI and MCP
  • Up to 3 reviewers on a review, on a link
  • 50 MB a file, and 30 days to recover expired media
Most teams

Team

For a team shipping agent-written changes every day.

$49per month

Start free, upgrade later
Evidence units
5,000 / month
Reviews
250 / month
Named members
5
Media retention
1 year
  • Everything in Free
  • 15 agent keys across 25 projects
  • Screenshots and video up to 30 min
  • Structured feedback over API, CLI and MCP
  • Up to 10 reviewers on a review, on a link

We are in early access, so accounts created before billing was switched on keep their existing limits. The price and every allowance are starting points we are still measuring against what this costs to run, and they will move before launch. If you hit one, tell us — we would rather move the number than break your loop.

Stop taking its word for it.

Create a workspace, make an agent token, paste the prompt into whatever is writing your code. The next thing it builds arrives with its own evidence attached.

Start free