Connect with the Product Hunt community

Nothing frustrates a customer more than repeating themselves on every call.

Memories lets your agent recall what happened in earlier conversations with the same person, so a returning caller does not start from zero.
In @Vaani Developer Platform you can choose how many past calls it remembers. Let one agent see the history of related agents, so the handoff between them feels continuous. also bring in your own customer details from your systems, such as plan, last order or account status from APIs

The result is calls that feel like a conversation with someone who knows you.

What should an agent remember about a returning customer?

A call ends and then what?
Someone has to copy the outcome into a spreadsheet, a CRM or a chat channel.

Custom webhooks send an update to the address you choose when something happens: a call starts, ends, is not answered, fails or is transferred.

• Choose the events you care about.

• Works with no-code tools like Zapier, Make and n8n.

• Updates carry the phone number, so you can match them to your customer records.
  All of this only on @Vaani Developer Platform

What should happen automatically after a call ends?

A brand can appear frequently in AI-generated answers, but the context of those mentions matters.

For example, ChatGPT might recommend a company as the best fit for small businesses, mention it as an alternative for enterprise buyers, or cite its website while recommending another vendor.

Lingo.dev can now automatically provision your localization engine s context from your website: generate a glossary, translation rules, and a brand voice, then save and attach them to the engine.

When creating an engine in Infrastructure, enter your website. Lingo.dev reads the content and creates the context automatically, ready for you to review and refine. Everything stays editable. Try a representative translation in Playground before using the engine in your integrations.

For an existing engine, select Fill missing context, enter your website, and choose which missing types to generate. Types already attached to the engine are left alone. Existing context isn t replaced.

A voice agent that makes up answers is worse than no agent at all. Callers ask about prices, policies and product details, and a wrong answer costs trust.

With Knowledge Base, you can upload your files, paste text, or point it to your website. During a call, the agent looks up the answer in your material and uses it.

• Works with long, multi-page documents.

• Add or update content without rewriting the agent.

• You can switch it off for agents that do not need lookups.

A short pause can happen while it looks something up. The answer is worth the wait.

I'm Shun, the maker of Ownfeed.

We launched here today, and I want to use this thread to learn from you, not to pitch.

We ran a Lovable checkout app through FetchSandbox. Six checks passed. Two stayed not checked.

We delayed an email response, but the webhook still returned success. We hadn't shown it recovering after a failed callback. So the overall run stayed incomplete.

That's the thing I want us to get right as we build this. Did the test cause the failure, or did the app just succeed? Same question for a background job that might crash halfway through.

What evidence do you ask for before trusting a recovery test?

Calling numbers on a Do Not Disturb list wastes money and can get your number blocked.

Before a campaign starts, DND check screens your contact list and removes registered numbers, so your agent only calls reachable people @Vaani Developer Platform

For test calls on your own phone you can skip it, but only use that on numbers you own or have permission to call.

Do you check DND before every campaign on @Vaani Developer Platform

This is something we've been working on at Vaani AI @Vaani Developer Platform . We wanted to solve a simple but important problem: a caller shouldn't feel like they're starting a new conversation every time the system switches agents.

The challenge is that passing a summary isn't always enough. Small but important details can get lost, like a callback time, a promise made earlier, the caller's language preference, or the context behind a decision.

At Vaani, we've been working on making agent handoffs more seamless, so the next agent can pick up the conversation without making the caller repeat themselves or creating a disconnected experience.

But there's still an interesting design question here. How much context should move between agents? Should every agent have access to the full conversation, or should each handoff carry only the relevant context?

Writing a good agent prompt is a skill most teams do not have.

Teacher AI is a chat where you tell your agent what to do in plain words, like "add insurance related points", and it updates the instructions. You see exactly what changed and approve it, so nothing happens behind your back. It asks a few questions when your request is unclear, instead of guessing. It starts from proven setups for common jobs like admissions, collections, reception, booking and lead qualification. You can upload a document in the chat to teach it faster.

What is harder for you: writing instructions, or testing them?

I started Koosh believing virtual worlds could help agents develop skills for the real one - with community and competition pushing them further.

Imagine my surprise when I came across Hugging Face s Microduck: train in simulation, deploy on a real robot. Different projects, a familiar mindset. Apparently, my train of thought had company. And a beak.

Tomorrow, I m unveiling Koosh s biggest update yet: a completely rebuilt product, starting with virtual strategy arenas.

Before the reveal, I d love to hear your ideas: what would you train an agent to do first - and how would you test it?

Many callers mix languages, for example Hindi and English, in the same call. Voice agents often get stuck in one language or flip back and forth.

Our agents follow the caller. If someone says "let's speak in English", the agent switches right away and stays there. You decide whether it should stick to one language or keep adapting.

Even the agent's own system phrases, like "are you still there?" and goodbyes, now match the language of the call.

What languages do your customers actually speak on calls?

A voice agent that can only talk is not enough. Customers want results: "Where is my order?" "Update my address."

Custom tools connect your agent to your own systems. You can set one up by pasting an example request from your API's docs, and the platform fills in the rest, including when the agent should use it and what it needs to ask the caller.

• Works with the services you already use.

• Pull information from the answer and use it later in the call.

• Control which agents can use which tools.

What is the first thing you would want your agent to do for you?

We re testing whether agents can discover a site s docs from its homepage, and what happens when API endpoints require authentication. Those are different problems, but both can make it hard for an agent to verify what a site supports.
How are you making your docs and APIs discoverable today? Do you link docs in the HTML the server sends, publish a public schema or read-only endpoint, or take another approach? What s worked for you?
We re building Dedica to check whether agents can access and use websites. If you re exploring this too, here s the page: https://www.producthunt.com/prod...

One thing we noticed while building Vaani is that the job doesn t really end when the call ends.

Clients want to know what happened, whether the agent actually completed its task, what data was captured and whether something went wrong.

We ve worked on this at Vaani and built a post-call processing flow that handles:

• Call summaries and structured data extraction.

• Success evaluation against the agent s actual goals.

• Tool-call and transfer analysis.

• Call outcomes and sentiment.

• Webhook delivery so clients can use the results in their own workflows.

Since launch, your bots and agents can check a change before they apply it.

Aperture is now an MCP server. Grok Bot, Claude Code, Cursor or any MCP client can call check_change before it applies a change. It checks that the files parse, the imports resolve, and the TypeScript compiler finds no new errors, including in files the agent never opened. It also flags a change that skips or cuts short the tests. Nothing the agent sends is run or kept. Setup takes three steps at https://aperturesais.grok.me/agents

Aperture Agent Check runs the same checks on every pull request, as a CLI or a GitHub Action. Red lines get annotations right on the diff. If a failure was already there in the base branch, it shows as amber, not blamed on the PR. Your code never leaves the runner. Aperture's own PRs go through it.

Honest numbers: in the benchmark, the editor catches 27 of 33 bad edits, with 1 false alarm in 15 good ones. check_change stops 20 of the 33 without running anything. The other 7 only show up when tests run, so keep running them. Every miss is listed at https://aperturesais.grok.me/ben...

Many callers mix languages, for example, Hindi and English, in the same call. Voice agents often get stuck in one language or flip back and forth between languages.

Our agents follow the caller. If someone says "let's speak in English", if the speaker switches languages, the agent switches right away and stays there. You, the builder, decide whether it should stick to one language or keep adapting.

Even the agent's own system phrases, like "Are you still there?" and goodbyes, now match the language of the call.

Tell me, does this feature speak to you?

Most data work isn t analysis. It s setup and upkeep.

Creating a data source for a new client. Pushing rows from an internal tool. Building the same custom metric for the fifth account in a row. Checking why last night s ingestion failed. Updating permissions when someone leaves the team.

All of it lives behind clicks. So it doesn t run in CI, it doesn t run on a schedule, and your AI agent can t touch it.

That s why we built Databox CLI - launching on Product Hunt on November 6. It puts the Databox API in your terminal: data sources, datasets, data ingestion, custom metrics, databoards, users and client accounts. One install, JSON or CSV output, clear exit codes for scripts. It also ships with agent skills, so Claude Code (or any agent that runs shell commands) can manage your Databox setup for you.

Thanks for following Comcent! We're putting out short tutorials, one task each.

This one: taking texts on a number you already have with Twilio or Telnyx. Tick SMS, choose who answers (a bot first, or straight to your team), and texts show up in the same inbox as calls, chat and email.

Video: https://youtu.be/L7cf2qbZ0n8

Really excited to take our best product yet with you and have best product people experience it first hand.
Avail 1000 platform credits, exclusive just for PH users by using the code PHL1000X.
website: https://www.vaani.dev/
Platform link: https://platform.vaani.dev/
The comments and this thread is open for everything about the product, thank you so much for trying out the platform!

A call ends and then what?
Someone has to copy the outcome into a spreadsheet, a CRM or a chat channel.

Custom webhooks send an update to the address you choose when something happens: a call starts, ends, is not answered, fails or is transferred.
Choose the events you care about.

Works with no-code tools like Zapier, Make and n8n.

Updates carry the phone number, so you can match them to your customer records.

I've been experimenting with giving AI access to our actual marketing data instead of just asking it generic SEO questions.

So I connected Google Analytics and Google Search Console to Claude using ZapDigits MCP and asked it to audit our website and find things we should fix.

The results were pretty interesting.

A few things it found:

we launched opencommerce today: one endpoint for your agent to search, check out and track orders across 5,000+ online stores. it has its own page, so following Agentcard won't show it to you: https://www.producthunt.com/prod...

try it and tell us what you think.

There s a growing tendency to flag AI-generated text. There s even a term for low-quality AI-generated content - AI slop.

And honestly, I understand why.

Communication should feel real. If every comment, post, or message is generated by AI, the picture becomes blurry.

But then I have a question.

On October 1, 2026, Google updated its guidance on AI-generated content. One line stands out:

"It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing."

Google Search Central, updated 1 October 2026.

That is not a suggestion. It is a requirement.