Two loops, feeding one memory.
Observant learns from your users in two ways: a standing 1:1 line with the feedback partners you invite, and light signals collected inside the product itself. Both feed one memory of each user — so what you know keeps deepening the longer it runs, and nobody gets asked the same thing twice.
A continuous 1:1 with the users you most want to hear from.
This is the active side. You invite the users whose feedback you most want; the ones who opt in become your feedback partners, each with a private line to your product on the channel they picked.
Start with your most-engaged users.
You choose who to ask. Whoever opts in becomes a feedback partner and picks where the conversation lives — email or Telegram. It's their own thread with your product, and they can step out in one tap.
Write what you want to learn, in your own words.
Drop a question the way you'd say it to a colleague. Observant makes the depth call — for a light topic, a couple of quick questions right in the channel; when it deserves more, an invitation to a roughly ten-minute AI-guided voice conversation. You write the intent; Observant phrases everything a partner actually sees.
voice, when the topic goes deepReplies come back as real threads.
Partners answer on their own time, in their own words, and each thread keeps its history. Answering your team's questions is what earns rewards, and they accrue with every answer. Volunteered feedback is always welcome and always read — it just isn't what the rewards are for.
Every partner builds a relationship memory.
The gist of who they are, the key points they've made, and a golden quote in their own words — carried into every later conversation. A follow-up becomes one rephrased question in context, and no one is ever asked to start from the beginning.
One line of code that listens while people use the product.
This is the passive side. A single snippet — added through a pull request you review, or pasted in yourself — collects the light, in-the-moment signals that tell you what's reacting right now.
It loads with one line, reads none of your codebase, and if it ever fails, it fails silently — your app never breaks because of it.
Thumbs up or down on AI outputs, quick satisfaction checks, an exit question the moment someone leaves or downgrades, and an always-there feedback button.
The product events you forward — what people actually did, sitting alongside what they said, so every reply has its behavior next to it.
Simple rules turn a spike into a question. Three thumbs-down in a day, and Observant surfaces a suggested question for your team to review before it joins the 1:1 loop. The in-product side spots it; the partners explain it.
One place where the learning lands.
Everything the two loops collect resolves into a portal your whole team can read — and ask.
Threads & memory
Every partner's live 1:1 thread, next to their relationship memory — who they are, what they've said, the quote worth keeping.
Insights & Ask Observant
Standing insight cards that form across all your feedback, plus Ask Observant when you have a specific question and want a grounded answer.
Digest & alerts
A weekly digest of what moved, and an alert right away when something shouldn't wait a week.
MCP server
Your coding agents and editors pull the learning mid-task, straight from the source. Read the docs →