Connect with the Product Hunt community

Coding agents understand your repository well. They know very little about the machine it runs on.

Ask one to free up disk space and it often reaches for rm -rf. Ask what is using port 3000 and it pipes raw lsof output into the conversation.

Termi Protocol is live on Product Hunt again today. This one is built almost entirely from what you told us after the first launch.

The headline is the limit handoff. When an agent hits a usage, rate or context limit in the middle of a task, Termi saves the session (goal, changed files, commands, errors, open todos and the conversation) and hands it to an agent on another model, which continues from the last finished step. Nothing is sent until you confirm.

Also new since July:

Google started enforcing app registration in four countries three weeks ago, and the rollout widens in 2027. Most Play apps were registered automatically, but "most" is not the same as "yours".

Google's Android Developers Blog (18 June 2026) set 30 September 2026 as the date developer verification protections begin, starting with users in Brazil, Indonesia, Singapore and Thailand, on certified Android devices. Google's developer verification page says it expands globally in 2027 and beyond.

As a growth designer, I look at newsletters from two perspectives: how fast they prove their value, and whether they make me feel like I m giving away my inbox for nothing.

Getting an email address is relatively easy. The harder part is designing a sign-up experience that promises genuine value without feeling like a bait-and-switch.

Going through 40+ opt-in pages recently, a few things stood out to me:

Most pages use vague promises like "Join 10,000 readers." A high-performing page tells me exactly what I will learn in the first issue.
The best opt-ins don't ask for a commitment upfront they offer a free sample issue or an archive link so I can test the waters first.
Trust matters: the pages that kept me engaged showed a clear frequency (e.g., "One email every Tuesday") and a visible "unsubscribe anytime" button.

I maintain Termexo. This AI-assisted update was checked against the 0.10.15 release notes.

This release is about finding the right project and making the next action clearer:

- Organize workspaces in nested folders, drag to reorder, and add tags and icons. Search the sidebar or open a project's directory from its menu.
- Choose Claude browser, Console or SSO sign-in, or Codex browser or device-code sign-in. The official CLI still handles authorization; its login terminal has a separate status.
- Discover models from a public directory and visible local Codex cache entries. Supported default profiles can show upgrade suggestions, which you review before applying. A catalog entry does not guarantee access through every account or gateway.
- Confirm before closing an active Agent terminal, so ending a process is an explicit choice.

Everyone has one. "Keep it short." "No emojis, ever." "My dog's name is Biscuit, stop calling him Buddy." The thing you've typed so many times you could say it in your sleep.

Here's the game:

1. Drop yours in the comments, exactly how you'd say it.

Building apps is my job, and AI is part of how I do it. After hours, I can't always tell if it's still work.

In Unprompt, I set my work hours once. It uses them to sort work from personal.

The blurry part is my own time: a side project at 10pm, a script for a friend on a Sunday.

How do you draw that line? Fixed hours, or a separate machine?

Genie, our AI Analyst, can already tell you what changed in your numbers. It's never been able to tell you why, because the why almost never lives in Databox.

Revenue drops 8%, and the real reason is a stalled deal in HubSpot, a pricing objection buried in a Slack thread, or a support ticket nobody flagged. Right now, finding that out means leaving your dashboard, opening three other tools, and piecing it together yourself. Every time.

Next week, we're launching MCP Connectors. Connect your CRM, Slack, support desk, or any MCP server, and your AI Analyst gets that same context, right inside the conversation. Ask why a number moved, and it points to the actual deal or thread behind it, not just the number.

Hey all, a few updates since last time:

Bring your own Anthropic, OpenAI, Google, or Mistral AI key for localization. Requests go directly to the provider on your key, and the provider bills you for those tokens.

Open Settings BYOK, add your provider key, and choose that provider s models in your localization engine.

Your key is used for the engine s translation and editing steps. Your integrations keep using its glossary, rules, and brand voice, while Lingo.dev runs the localization infrastructure.

DungeonQ is a defensive deception runtime for suspicious AI-agent sessions. A trusted gateway routes a designated session into a persistent synthetic world, with useful world-only Wrong Tickets, observable activity and separately approved bounded changes. The goal is to give people and authorized defender agents room to act.

v0.16 addresses a concrete lesson: protecting one route is not enough.

In the earlier Sol experiment, all three diverted sessions received the artificial target through an open alternate source, A2, despite zero reads through the main source route. Those failures remain published.

The repair applies shared tenant, resource and credential-version checks to the main source, alternate source, cache and export. A2 keeps no target copy. Cached content requires a fresh authority check; an unavailable origin produces UNKNOWN instead of stale disclosure. A world ticket or claimed role cannot grant source access.

I m getting closer to launching my tool on Product Hunt, hopefully sometime in the second half of this year.

I currently have a 3-phase launch strategy in mind.

Phase 1 Free version

Every Mac with Apple Silicon has a Neural Engine. More apps use it every month for dictation, transcription, and other AI that runs on the Mac.

Activity Monitor doesn't show it. I had no way to see how busy it was or which app was using it.

rhun is a small, open-source code editor. It has extensions now, and I d love to see what people make with them.

We ve added Claude Code and Codex skills for building and testing your own extension, plus a skill for porting open-source VS Code extensions. Describe the tool you want, then work on it with your agent. Rust/Go templates, examples and CLI commands give you a place to start and a way to share it.

What s the oddly specific extension you ve always wanted?

Try rhun: https://rhun.app

I run 10+ products solo and read the support inboxes myself. After a press release for one of them, the contact address got buried in pitches: webinar invites, free whitepapers, agencies wanting to run my ads. Gmail filters didn't hold. Senders change every time, and "thank you for your inquiry" shows up in both auto-replies and real human replies.

The worst ones borrow a platform. Someone shares a pitch as a Google Doc, so the email comes from Google. Or they sign you up for their Zoom webinar, so the invite comes from Zoom. You can't send Google or Zoom to spam.

Jev is what finally fixed it. It doesn't write anything, it just tells you how likely each yes/no answer is, and you decide where the line goes. Per email I ask five yes/no questions in one call: did a person write this, should we reply, are they selling us something, do they want a meeting, are they clearly asking us to stop. A few rules around it:

Ask twice, act only if both agree. Otherwise I look myself.

Hey Product Hunt! I'm Kristina, working on marketing, back with another product after some time away.

I know from experience how hard it can be to get back into job hunting and interviews after a long time in one role. It's like learning a new skill.

Hi everyone

Thanks for following Cloud Releases.

Quick update from me: I've just launched a second tool today (Oct 16), and I think a lot of you will find it useful.

AgentDock 1.4 is out. The short version: it now works on a Mac without cmux, tells you what your agents tokens would cost at API prices, lets you leave a message for a busy agent, answers questions from other agents over MCP, and reads seven more agents.

Works without cmux

• If you run Claude Code or Codex in any terminal, or in the Claude or Codex app, AgentDock now shows them on a Mac that doesn t have cmux. Before, the strip and the menu showed only the cmux error. cmux is optional.

Costs at API prices

When I launched Panza on Product Hunt a few days ago, I had pretty much zero expectations.

No big campaign, no coordinated launch effort. I just put my little indie app out there to see what would happen.

70 upvotes later, I m still a little blown away.

Hi everyone,

In May we launched Bond here and finished #1 Product of the Week.

Hi everyone,

In April we launched Ovren here and finished #4 Product of the Day. The question we heard most since: "The change is ready, but how do I know it looks right before it goes live?"

On October 22 we're launching Live Preview.

Describe the change you want in plain words, like "add a customer reviews section under the banner." Ovren makes the change on a safe copy of your website or app and shows you the result: your real page, with the change on it. Click around it like a visitor would. If it looks right, approve it and your team publishes it. If not, send it back with a note.

On 2 October Apple said it will make granting Full Disk Access on macOS require more explicit user action, and it named AI agents as the reason. If you let agents touch your files, that's a prompt to audit what you've already handed over.

On 2 October Apple posted a short developer news item titled "Updates to Full Disk Access in macOS." It says Apple will add controls that require more explicit user action when Full Disk Access is granted, because some developers misuse it to reach sensitive data such as files, mail, messages and browsing history without the user fully understanding. The sentence I keep coming back to is this one: "As AI agents become increasingly capable and autonomous, the risks associated with this level of access will grow substantially."

Big news for Sendr!

Hey makers

NotchDodo now keeps your Mac awake while your AI agent works, then puts it to sleep once the job is done and you're away.

I normally keep 8 or more sessions of Claude Code open simultaneously. At some point, I discovered each new session was launching a complete copy of the MCP servers I've set up so far (Gmail, Calendar, Drive, Zoom, LINE and one more). Each one required approximately 560 megabytes, so my computer was carrying multiple copies of these processes at all times, all through the day, even when nothing was calling them.

Thus, I decided to get rid of them from the always-on list and start them only when necessary. Now I only need to execute a small command (mcp-call) to get a particular tool running. For instance, running a search on my emails is simply mcp-call gmail-local with the search tool and a query, and the process finishes right after that. The agent can still find the tools since a short skill lists them, and the hooks that used to check the old MCP calls (the HTML format of emails, asking before anything gets sent or deleted) now run on the new command too. I actually thought it was likely that at least one of these would break.