跳到正文
Modem:Blog·· 8 小时前AI 评分44

Modem 上线应用内反馈表单:用户提交即可触发智能体自动建功能

Add a feedback form that builds what users asked for

AI 导读

Modem 向所有用户开放应用内反馈功能,可在 Settings > Integrations > In-app feedback 配置,或把起始提示词粘贴给编码智能体自动接入。

正文 · 原文

In-app feedback is now available for all Modem users. Set it up from Settings > Integrations > In-app feedback, or paste the starter prompt into your coding agent.

At most software companies, user feedback arrives by email or in a ticketing tool. Usually the support team reads it first. Then maybe (maybe) someone escalates it to the product team to take a look. By the time an engineer sees the request, days or even weeks have gone by, and the user has moved on.

What if you could act on feedback the moment a user hits "send"? With Modem, you can add a feedback form to your app that kicks off an agent workflow as soon as a request comes in. Have it file a Linear issue, ping the team that owns that part of your product in Slack, and yes, even auto-build the feature.

A form that belongs to you

Modem's feedback form is a shadcn/ui component: a feedback dialog with screenshot capture, plus a crash report form for your error boundary. The source is copied into your repo, so you can adapt it to your design system like any component you wrote yourself.

npx shadcn@latest add https://app.modem.dev/r/feedback.json

Or let your coding agent do it. Paste the starter prompt from Settings into your favorite coding agent (connected to Modem's MCP server), and Modem's install-feedback skill creates a feedback channel, stores its token, adds the form, and sends a test submission.

Modem's feedback dialog shown in four design systems: the shadcn/ui default, a dark theme, a brutalist theme, and an editorial theme. The fields and copy stay the same; only the colors, corners and type change.

Modem's feedback form matches any app's look and feel

There are no third-party scripts or externally hosted code. The form runs entirely inside your app, and nothing leaves your user's browser until they hit send. Then the submission goes to Modem, just like the other channels it monitors: Slack Connect, help desk tools, and call transcripts.

What happens when a user hits send

Modem matches the sender to a person and company, then files the submission on a topic with every other report of the same problem. It doesn't matter where the other reports came from.

Let's say a user submits a request for "faster exports". If other users requested it on other channels (like Slack Connect or Zendesk), they get grouped together automatically.

A user reports an export timeout. Modem files it on the existing topic, and an automation posts it to #product with a suggested owner.

Grouping user requests together is what makes Modem special. It means you can trigger automations once two or three or more users have requested the same thing. Or do it the moment you see something the first time. It's up to you.

Hand the request to a coding agent

Every form submission fires a Feedback Activity event inside Modem. It carries the topic the submission joined and all the related activity: who asked for it, when, how many times, and more.

With smart triggers, you can set up automations that act on plain-English conditions like "when three or more users ask for this feature". Jev, the decision model behind smart triggers, checks the condition before anything runs. From there you can create tickets, ping teammates, or just build the feature outright with one of our coding agent integrations.

When

a feedback submission has new activity

and Jev decides

the submission is a feature request, and its topic has three or more reports

Then the agent

Starts Claude Code on a first pass, with each user's request and screenshots as context, then posts the PR to #product.

Getting this running is easy. Just ask the Modem agent in the app, in Slack, or from your own Modem-connected coding agent:

When three or more users request the same feature, have Claude Code build a first pass and post the PR to #product.

Modem makes it easy to auto-build features as they come in through your feedback form, or from any support channel it monitors. Claude Code, Cursor, and Devin all work, with more platforms being added all the time.

Follow up when it ships

Shipping the feature is the easy part. Remembering who asked for it is harder.

When the PR merges, Modem posts to Slack with the name and company of everyone who asked for it, whether they used the form, a support ticket, or a shared channel. Turn it on with the Close the loop template in the automation library.

Build the thing, then tell the user who asked for it.

Other ways to collect feedback

The web form isn't the only way in. Modem takes feedback from two other places, too.

Your product's AI assistant. Connect it to Modem's hosted feedback MCP server at https://mcp.modem.dev/feedback. When one of the assistant's tools times out or returns something it can't use, the assistant offers to report the problem, and sends the report only if the user says yes.

A CLI, mobile app, or backend. POST submissions to the feedback endpoint with the same channel token.

Either way, the reports land on the same topics as everything else and can trigger the same automations.

Get started

  1. Go to Settings > Integrations > In-app feedback and click Copy prompt
  2. Paste it into your coding agent with your project open, then deploy with the new environment variable
  3. Connect Claude Code, Cursor, or Devin from the coding agents settings
  4. Ask Modem to create the automation above, and add Close the loop from the automation templates

The in-app feedback docs cover tokens, allowed origins, the endpoint, and the MCP server. New to Modem? Sign up on the free plan and start with the prompt.

来源:Modem:Blog · modem.ai