Note: you linked the sign-in page, not cominty.ai. Include it if you want landing page feedback.
1. What did you build? Did you succeed?
An Instagram agent (answer DMs/comments, DM people who comment on a post), then a HeyReach LinkedIn responder. Creating both worked, but I got stuck at deploy: I'd have to do the integration myself. I was just trying it out, and that's your challenge: people will try it because it looks great, then drop off once it gets complicated.
2. What impressed you?
- Onboarding: really good, and starting people in "What do you want to build?" is smart.
- Create Agent button: clear and simple. It picked Sonnet for my agent, exactly right, and model switching is a real value point.
- "Copy prompt": paste into your coding agent and done. Love it.
- Templates (support, lead researcher, document export): exactly what's needed.
- Smooth OTP signup, auto favicon, solid docs.
3. What was frustrating?
- Test output: I said "just send me the candidates" and it replied this seemed part of an earlier conversation, not like replying to a LinkedIn contact. Mostly not your fault, but my instructions were weak and nothing pushed me to improve them.
- Clicking the sidebar kills a running chat.
- No "Edit with AI" after creating. Add it next to the agent name and in the three-dot menu.
- "Create with AI" chat box looks like a plain text box, so people won't invest in it, and pasting long text breaks it. Make Create Agent the main flow. That page is really amazing.
- Skip in onboarding throws you out instantly (add a confirm). Get Started is hard to find after (move it to the top bar) and stays at 1 of 4 without an API key.
- Small: input box should auto-grow, expired OTP still opens the OTP page, show local currency, very desktop-heavy.
4. How does it compare?
Haven't used Lindy, Relevance or n8n agents. Zapier: expensive and not great. Cominty is way simpler. But right now it feels like the OpenAI playground with a nicer UI, so why wouldn't a dev just use OpenAI? Pick an angle: "easy" (then integration must be easy too) or "advanced" (then show a clear different value).
5. What's missing?
- Instruction guidance: a personality field, do's and don'ts, testing tips. Push people to set it up properly, or they'll get bad output and blame you.
- Realistic test mode: simulate an incoming LinkedIn message, not a random chat.
- The last mile: retrieve data (scheduled or webhook) → act (draft replies) → send back (optionally with approval). Basically OpenAI + n8n. Frame setup as: What comes in? What should it do? What goes out?
- Social connectors: Instagram and LinkedIn DMs. I built a LinkedIn responder with HeyReach + MCP before and haven't seen anyone solve it well. Happy to share my repo.
6. Would I use it for real?
Not yet, I'd still integrate it myself. Close that gap and I'd use it for my LinkedIn responder. Pricing is fair: a few agents and messages cost 8 cents. People will pay good money for a solution like this.
7. What to improve first?
Examples. Show what you and others built, let people share agents, and go beyond support bots. Also add a personal touch: why you built it, how you use it. TestMyBuild worked because I solved my own simple problem. Solve one problem really well.
Challenge: build a LinkedIn DM agent (via HeyReach) entirely inside Cominty. You'll see exactly what would make this platform flawless.
Overall: tough industry, great base. Close the integration gap, guide better instructions, add examples, and this becomes really useful.

