Feedback → your assistant → an approved plan

Your users show you. Your AI helps you decide.

One suggestion, followed through

Example workspace

FieldnotesProjects

No projects yet1

A user’s suggestion

“I opened it and didn’t know where to begin.”

Screenshot + pin 1

Someone points to the part of your product that feels confusing.

Illustrated example · No data is sent or product changed

Keep the original suggestion and screenshot together. Ask your assistant what matters, what fits your product, and what to improve next.

Request free beta access

By invitation. Already using Patchkind? Open AI connections.

See the actual product

From a note to an approved brief. In 18 seconds.

Watch a demonstration project collect a suggestion and screenshot, then prepare a brief for a coding agent.

Recorded interface; some styling has since changed. This shows collection and approval, not automatic coding.

Real browser captures · Demonstration project · Silent

Read the video description

A person suggests a product walkthrough. Their note and screenshot reach the owner together. The owner checks the scope and expected result, then approves a brief for their coding agent. The owner still checks and publishes the finished change.

Start with one useful conversation.

Set up your product. In your invited workspace, create a project, write its direction and install the feedback button using Connect. See the website installation guide.
Connect your assistant. Open Account & settings → AI connections. Use the Patchkind plugin when it is available in your account, or the listed connection instructions for a compatible MCP client. Choose read-only access if you only want analysis. Never paste an access key into a chat.
Choose the project and review the evidence. Use the prompt below. The assistant should tell you what it read, show the supporting feedback and acknowledge missing screenshots.
Approve, build, then verify. With plan permissions, the assistant can save a plan and record your approval of that exact version. Your coding agent still needs project access. Review its changes and tests before publishing through your existing process.

Your first conversation

Use Patchkind to review my product feedback. First show my projects so I can choose the right one. Read all feedback for that project and inspect the relevant screenshots and pins. Compare suggestions with my product direction. Separate repeated requests from one-off opinions and explain what you could not inspect. Recommend one small improvement and show its scope, evidence and checks. Do not approve work or change code until I approve the exact plan.

Know what you’re connecting.

The connection lets your assistant read feedback belonging to your account, including requested private screenshots. It does not give anyone public access to those images. Your assistant’s provider processes the material it receives. Read the data details and controls.

You can use your own assistant without enabling Patchkind’s built-in dashboard AI. Built-in AI review is not enabled in this beta. Recordings can be downloaded, but the connection does not analyze video.

The plugin has been tested with the service owner’s account. Public directory availability and a separate customer’s connection are not yet verified. If you cannot find it, contact the team; an image-inclusive brief download remains available.

Connecting does not start an unattended worker, merge code or deploy your product. An implementation result is not proof of a live release. After you verify and release a change, its reporter can tell you whether it helped.