PH热榜 | 2026-08-12

@pritesh_kumar3 Congrats. What’s the most realistic self-host setup you’ve seen work in production; single server, Docker, Kubernetes? And roughly what infra cost should a small agency expect to run 500–1,000 call minutes/day end-to-end on their own stack?

@pritesh_kumar3 Congrats on the launch! I can't wait to use it to develop my very own voice bot.

@nuseir_yassin1 - Thanks for your message.

We have very excellent support for human handoff, where you can declare static destinations or destinations based on the context from the conversation of the agent with the caller. You can define PSTN phone numbers or SIP addresses or ViciDial Ingroups in those destinations where the voice agent can transfer the call based on your matching conditions.

And, of course, have to prompt the LLM to make that transfer call based on certain conditions. If prompted well, the LLM can do it with very high confidence.

Hello@nuseir_yassin1 thank you for your message. Adding to Abhishek's reply - the failure path is worth noting too: if the transfer destination can't be resolved (resolver timeout, no match, etc.), it fails gracefully and the agent keeps the conversation pipeline running rather than dropping the call. And for tightening when it hands off, the QA node surfaces signals like repeats, interruptions, and dead air that teams use to refine the confidence logic over time.

@nuseir_yassin1 Dograh can hand off to static or dynamically resolved destinations, including phone numbers, SIP addresses, and ViciDial ingroups. The agent dials the destination and waits for an answer before bridging the caller, with configurable transition messages, timeouts, and fallback behavior if the transfer fails.

@lucianb - Thanks, and good question. Sentiment scores are easy to compute but hard to act on mid-call unless we have super robust orchestrator spanning multiple LLMs of different sizes and capabilities (something we are working on).

The signals that actually predict a bad conversation are structural and post call: how often the caller repeats themselves, how often they interrupt, how long the silence runs before they answer (dead air), whether they ask for a human and refuse to talk to the agent. We handle this using a native QA node in workflow builder and also integrate with solutions like Tuner, Noveum. On quality we hold one metric above the others: did the caller get what they called for. Everything else is a proxy for it.

Privacy is where open source does the real work. Self-hosted, the audio never leaves your infrastructure, so there is no third party to trust with it. You keep the derived signals and drop the recording.

@lucianb Hey Lucian, thanks for the question. We have an inbuilt QA node that you can run across 100% or a smaller sample of your live calls. The QA node has inbuilt functionality to do sentiment analysis or any custom analysis for your use case.
And since dograh is self hostable you can add a guardrail/mask layer to send PII anonymised data- though this might still need some plumbing and not available out of the box.
Hope this helps :)

@lucianb Thanks for the thoughtful question. We focus on practical call outcomes while giving teams control over how their data is processed, stored, and retained.

@ankushkun Thank you for your query. We actually have first class support for multi step conversation. When the entire industry was doing one monolithic prompt for voice agent, we and few others (like Bland) were doing natively multi node agents which broke a big monolithic prompt in multi step conversation.

This was specially challenging with models like Gemini Realtime which does not allow updating agent prompt in between conversation, but we figure out some hacks which allows our multi step conversation to work well with them (reconnecting with context playback and compaction).

Hi @ankushkun we support multi step conversations extensively.
Infact with Dograh you can build multi-step conversations for real time/speech-to-speech models as well (apart form cascade)

@ankushkun Yes, Dograh supports multi-step conversations and customizable workflows for different use cases.

@ritesh2503 - Thanks for your appreciation. Yes, we as builders are super passionate about what we are building for the open source ecosystem and community. Our mission is to enable every company in the world to own and operate their voice AI agents securely and efficiently.

@ritesh2503 Thanks for the support! Glad the self-hosted approach worked for your team at Floworks. Excited to keep building with feedback from teams using Dograh in production.

Thanks Ritesh- we believe in self hosted voice ai and also open weight models. We will keep building for the dev community :)

@ankitnayan - Thanks for your comment. Speech to speech are excellent models which offer language switching and lower latency. Many labs are working on Speech to Speech models, like @Google (Gemini Live), @OpenAI (OpenAI Live), @Grok (Grok AI) etc. @Dograh support these and many more.

The only challenge currently is that you cant use custom cloned voice with them, and the tool calling can sometimes be a bit unpredictable. However, we are seeing some very rapid progress in the space and these are bound to improve in short term.

@ankitnayan Mostly control. Speech-to-speech is faster and handles language switching well, but you currently give up custom cloned voices and some predictability around tool calls. The gap is closing quickly, though.

@ankitnayan While we have seen S2S win on latency and 'humanness'. Having said that tool calling and instruction following is not upto mark with LLM's in cascade approach. However with

@Dograh we ship the ability to break S2S stream accross multiple agents instead of 1 end to end stream and thus give you minute control over your instructions - shorter context and minimal instruciton following - thus enabling premium grade performance with S2S models.
We will be weleasing benchmarks soon . Keep an eye out :)

@brucem80 This means a lot, genuinely. Thanks for sticking with us these past few months :)

@brucem80 Thanks Bruce! Glad to hear Dograh and our support have been useful.

@brucem80 - Thanks a lot for your kind words ❤️

You will always find us here trying to evangelise and promote Open Source voice AI adoption. You will always find help on our Slack Community.

@hannesh - Thank you for your message.

We have given a lot of thought to how call transfer is configured and works on @Dograh. You can not only set up static destinations but also resolve dynamic destinations based on the call context. The experience is very seamless for the caller. We have the option to provide a transition message - like "Please hold on while I transfer your call to a live agent", and if anything breaks during the transfer, we have a very graceful fallback also which can be controlled by the prompt.

@hannesh thanks! from the caller's side it's pretty smooth - they hear hold music while we dial the destination in the background, and you can configure a heads-up message first so it's not abrupt ('connecting you to our billing team now'). importantly, we don't just throw the call over the wall - Dograh actually dials the destination and waits for it to answer before bridging the caller in, so they're never dropped onto a dead or unavailable line. there's also a configurable timeout in case nobody picks up. the one gap today is context isn't automatically passed to the human once connected, but that's very much in our roadmap.

@hannesh The transfer is seamless for the caller. Dograh plays hold music while dialing the destination and only bridges the call once someone answers. You can also configure a short transition message, timeout, and fallback behavior.

@shantanusewu - Thanks a lot for your support. ❤️

Please do not hesitate to reach out if you need any support in implementing Voice AI for your use cases.

Hi @shantanusewu thats such a lovely remark. Would love to connect on other platforms as well . we love IITD :)
Here's lin: https://www.linkedin.com/in/priteshkr/

@shantanusewu Thanks for the support! Really appreciate you rooting for Dograh and the founders. ❤️

@seomaxtech Thank you for your message.

Our cloud offering is priced at 1 cent per minute of calling if you bring your own keys for models compared to around 5 cents per minute from Vapi and Retell.

If you use Dograh managed models, our model usage is priced at around 7 cents per minute compared to around 8 - 12 cents per minute from Vapi and Retell.

@seomaxtech Our cloud offering costs $0.01 per minute with your own model keys, compared with around $0.05 per minute for Vapi and Retell. With Dograh-managed models, it is around $0.07 per minute, compared with roughly $0.08 to $0.12 per minute. And we can go much lower with an increase in volume.

@seomaxtech to summarise what the team said - self-host it and it's free forever, or use our cloud at 1¢/min + model usage. Your call on ownership vs. convenience :)

@drag0s - Thanks a lot for your kind words ❤️ Wishing you loads of luck with @Zaplingo

@drag0s Thanks for the support! Glad the local model support and hands-on help have been useful.

@drag0s the founder shoutout is so real, can confirm from the inside :) and it's not just the founders either - our Slack community's got contributors and users constantly trading configuration insights, figuring out the right tool setup for specific use cases, and helping each other debug.

@henry_habib - Thank you for your comment. Appreciate it. Good luck with @Voquill as well. ❤️

Thats core to our philosophy as well@henry_habib 

Thanks @henry_habib self hosting and OSS is core to our belief system. And thats how the world should be - accessible and open :)

@huisong_li - Thank you so much for your comment. Good luck with @HarnessRouter ❤️

Thanks for the support  @huisong_li 

thanks @huisong_li for your support :)

@aj_123 - Thanks a lot for your support.

Wishing you loads of success and luck with @Tough Tongue AI

Thanks @aj_123 . lovely to see you rooting for us

@aj_123 Thanks for the support! 🙌

@roopreddy - Thank you so much for your message.

Yes, for sure. We not only provided an inbuilt QA node, where you can customise the QA prompt for different industries, call types, and use cases, but we also integrate and play well with other vendors in the space, like Tuner and Noveum.

You can also expose those QA results in your post-call data sync (webhook nodes) so that your systems immediately get updated with how did the call go and how it can be improved.

Hook that with an MCP, and you have got a self-improving agent. All the best with @Documentation.AI

@roopreddy Yes, teams can customize the QA prompts and scoring criteria for different industries, call types, and use cases. Results can also be synced to external systems through post-call webhooks.

@abod_rehman Thank you for your message.

We are developers at heart, and we are creating this product for the developer community and business owners who want to own their voice AI stack.

Wishing you loads of luck with @Triforce Todos

Thanks @abod_rehman . we are obsessed with dev and oss ecosystem and are all devs ourselves as well

@abod_rehman Thanks, Abdul! We wanted developers to have the flexibility to choose their own models without being locked into one provider.

@devansh_pawan1 - Thank you so much for your support. You guys have done excellent work with @Smallest.ai and we are very happy and proud partners.

@devansh_pawan1 Thanks for the support! Great to be partnering with @Smallest.ai.

@dipanshu_kushwaha5 Thanks for your message.

Yes. 100%. Dograh is to voice agents what @n8n is to workflow automation. You can either decide to visually build voice agent using Dograh UI or you can use MCP tools offered by Dograh (cloud or self hosted) to talk to your coding agents to build an agent for you. And all of these play really well with various telephony providers so you can go to production with least friction.

@dipanshu_kushwaha Yes, you can build voice agents visually in Dograh without writing code. You can also use Dograh’s MCP tools with coding agents if you prefer.

@dipanshu_kushwaha5 To add one more layer - it's no-code even after you build. There's a Test Chat mode where you can edit or replay any turn in a past conversation and Dograh regenerates the agent's replies and node transitions from that point, so you can debug and refine logic without touching code. Makes it easy for non-technical folks (support/ops teams) to actually own the agent long-term, not just the initial build.

@nihalkumar_dwivedi Yes, agencies and developers can white-label Dograh, deploy it for customers, and charge under their own pricing model.

@nihalkumar_dwivedi - Thank you for your message.

Yes, @Dograh comes with a very permissive licence that allows you to self host and white-label the solution and use it for commercial purposes.

You can also use our hosted platform as a headless API provider if you have smaller volume to start with and then move on to self host on your own infrastructure as the volume grows.

@bobaikato - Thank you for your comment.

Inbound volume needs to scale more dynamically and is not in operator's control while outbound volume can be controlled by the operator.

Example: If there is a big event in town, a car dealership might get many more calls than they usually do on a normal days.

So, when it comes to scaling, the inbound needs to be more elastic and responsive, while the scaling requirements for outbound can be preplanned and provisioned.

Thats an excellent question @bobaikato 
Its mostly around agent building.
In general more receptive about AI agents in inbound calls( e.g. Customer support etc) rather than outbound calls (e.g. Insurance reminder etc)
In outbound we have seen that obsession around getting the first 15 seconds right works best. While in inbound scoping and being abl to handle variety of queries and handoff ecomes critical.

@bobaikato Inbound requires elastic capacity because call volume can spike unexpectedly. Outbound is easier to forecast and provision since the operator controls the call volume.

@aditi_gupta38 - Thank you for your comment.

When you sign up on Dograh Cloud or setup a Dograh stack in your self deployed infrastructure, you get free credits to get started and build your voice agents.

Hi @aditi_gupta38 You can signup at app.dograh.com . We set you up with free credits- both for hte platform as well as AI models.

@aditi_gupta38 Yes, you’ll receive free credits when you sign up for Dograh Cloud

@chilarai - We ship with an MCP server that can help you build an agent on Dograh. While we dont expect you to bring in earlier conversation recording or custom data, if you have them, the coding agent (using Dograh's MCP) can extract signals and context from there and be able to create an agent which works better for your use case.

You can always find more help on our Slack community.

@chilarai its ready from the go - you can build and test any voice agent for your use ccase - within 90 seconds (90sec is our SLA).
Give it a spin here: app.dograh.com

@chilarai It’s ready to use out of the box, so no custom training data is required. You can still connect your own data and customize the agent for your use case.

@kate_ramakaieva - Thank you for your message.

Yes, observability and automatic evals creation is something that's on top of our head. We do integrate our basic observability using OTEL exporters on Langfuse, where you can already create data sets for your own use cases.

We are also trying to add these features natively on Dograh and MCP so that observability around latency, token uses, and model performance becomes a first-class citizen of the platform on both Cloud and your self-hosted environments.

Wishing you all the best with @FuseBase

@kate_ramakaieva Thanks for the feedback. We’re working toward native observability for latency, token usage, model performance, and call quality across both cloud and self-hosted deployments.

@isroiljon - Thank you so much for your kind words.

We have given a lot of thought to creating the product and making it easy for developers and business owners alike to create and manage their voice agents. We are always hungry for more product feedback on how we can improve it and make it easier.

Wishing you all the best with @Premation

@isroiljon Appreciate you checking it out! Would love to hear what you build with Dograh.

thanks @isroiljon . do try our MCP's - we try to give a blowout experience to devs

@kasaei - Thank you so much for your kind words. We are always hungry for feedback. All the very best with @Api Hunt

@kumar_gautam Thank you for your message.

These are some very relevant questions. We connect over Websocket for TTS and STT, so any failure over websocket connection is automatically retried. For LLMs, we have fallbacks in place, so that if our primary LLM takes longer to respond or fails to respond, there are fallback LLMs in place.

I welcome you to try out @Dograh

@dzaitzow - Thank you so much for the kind words.

We strongly believe that once Dograh is into consideration of every company in the world for their voice AI use case, revenue will be an organic side effect of that. The primary focus right now is to be able to create as much value as possible for our early believers and customers to cement our place in this industry and market.

Enterprise tickets are of course something thats super welcome and helps us bring the revenue numbers up, and some of very large enterprises (Bn $$+) have told us that they are self hosting Dograh for their use case.

@yahia_bakour3 - Thank you so much for your support.

Wishing you loads of luck with @Context.dev

@rahul_aluri - Thank you for your message.

We are seeing good traction in Legal Intakes (inbound and outbound), Car Rentals (inbound), Restaurant Booking (inbound) and Medical Insurance (outbound) sectors.

The limitations are mostly around declaring about robo call for automated calls. Supporting regional languages are more of a capability concern and using the right set of models behind the orchestrator.

All the very best with @Spur.fit

@benln The website you linked goes to a twitter post? Im not logged into twitter so i cant see it. I found this website tho. https://x.ai/bot

@benln 🤌

@rrhoover I was surprised at how fast this spread internally. I don't think I've ever seen a product catch on so fast across so many different teams/use cases.

@alina_petrova3 Runs in parallel, but depends on how you prompt them! If you give a one Bot four tasks in a row it will keep context and work through them in sequence (but can spin up sub-agents or call another Bot to parallelize work if prompted to do so).

For parallelizing work without additional prompt context we recommend having a "chief of staff" Bot (or putting a few Bots in a group chat), and they will divide and conquer tasks and parallelize work as a default.

@conduit_design A bit different from PX Computer, each Grok Bot account gets a single persistent cloud machine (browser, files, terminal, live logins) that their Bots share, and Bots can coordinate in threads / group chats independently so you’re not the management layer between specialists!

Team-based control plane is what we're interested in in the future. This may start with "shared Bots" and starter packs, and then move towards fuller shared controls. You can join an enterprise waitlist to get early access to what this looks like :): https://cursor.com/contact-sales?product=grok-bot

@conduit_design On logins, there s user choice on what to use with each tool. In general, we:

1. Prefer high quality connectors/MCP where they exist

2. Can also have Bots use the real UI on the cloud computer

Login, 2FA, captcha, and payment credentials and hands the computer back to you to complete sensitive steps, and then sessions persist after that.

TLDR; session one can feel a bit like a "new device” (captchas, email codes, “is this you?”). After the first sign-in, friction drops because the session sticks on that machine. We've been running this internally and with a beta program of hundreds of testers for the past several weeks and haven't run any issues on sessions sticking. Overall it's been really fun to work with!

@nrique You could create an agent to help brainstorm post ideas, track engagement, and iterate on new ideas over time.

@pregasen Outstanding launch, excited to see Lettertrace quickly become the go-to standard for measuring brand visibility across LLMs, qq is there support for scheduled cron triggers built into the CLI, or is it intended to be hooked into existing GitHub Actions / CI pipelines?

@pregasen Congrats on the launch Mathew. Great vid....lead with top 5 hopefully. Good on you for open sourcing.

@denitsapenchevavaltchanova Awesome Grok and Meta AI is on next week's roadmap. We'll add!!

@mateuszkonik appreciate it :)

@hamza_afzal_butt Really appreciate the support 🙌

@hamza_afzal_butt thank you!

@gabriella_anjani thank you!!!

@milescward Exactly! That’s the gap we’re trying to fill. AI visibility is becoming an important part of marketing, but the analytics around it are still pretty fragmented. Really glad

@milescward thank you!

@meet_patel24 Really appreciate you giving it a try! Glad the BYOK approach made sense to you. We wanted people to actually test it out and see the value before having to pay for another tool. Thanks for the kind words

@meet_patel24 Thank you for the kind words!

@suryansh_tiwari2 That’s exactly what we’re hoping for. AI visibility is still pretty new, so we’re excited to see how teams end up using Lettertrace in their day-to-day workflows. Really appreciate the support!

@suryansh_tiwari2 thank you for the kind words + for sharing it on Linkedin!

@alina_petrova3 yep that's built in

@ben_futoriansky Thank you :))

@ben_futoriansky thank you Ben!! We love work with ya'll

@ben_futoriansky thanks so much Ben!

@seomaxtech hahaha i read you.

@dhanushreddy29 We could make that happen. @cmillstein could you help with that?

@tigran_chakhalyan1 thank you so much !

@riya_jawandhiya the rest of our stack helps with that! https://letter.company

@abdullah_javaid3 Yep, these are diferrent things. We can detect it semantically. How often it's cited, how often it's mentioned, and the sentiment on the mention itself!

@busmark_w_nika Watch the video :) https://letter.company

@madalina_barbu Thanks a lot!

@madalina_barbu thank you !!!

@benln  @marlonmisra Useful templates the document collector and client onboarding tools sound especially practical for service firms

@benln  @marlonmisra Congrats on the launch, team! 🚀

Solving that "last mile" problem between building a quick prototype in Claude Code and actually shipping a secure, client-ready portal with auth and custom branding is huge. The friction usually kills so many great custom workflows before clients ever see them.

Really impressed by the template library—the Document Collector and Project Tracker look like immediate wins for service firms.

Quick question: How are you handling permissions when clients need multi-tiered access (e.g., a manager vs. an end client)? Wishing you a massive launch day!

@benln  @marlonmisra The last mile is usually where these AI app builders get interesting.

Getting an AI to generate a working prototype is becoming almost trivial. Getting that prototype into a real client workflow with authentication, permissions, branding, and the right user context is a very different problem.

I like that Assembly is attacking that specific gap instead of trying to be another general purpose app builder. The fact that client identity and permissions are built into the platform makes the “build something for my clients” use case much more compelling.

200+ apps in private beta is a pretty solid signal too. Curious to see what firms end up replacing first.

@jordan_wechsler1 Your templates QA was impeccable!

@adam_schwartz7 your product work was (and IS) 🔥

@german_merlo1 thanks! we think Lovable is great BTW and many of our users use both of us in concert. A very common setup is Lovable for the marketing site and Assembly for the secure client experience.

@elene_tandashvili thank you!!

@zaczuo Hey Zac,

Awesome find with Unsloth Desktop on Product Hunt today! Running and training local AI models offline is a massive win for devs.

I'm an AI CRO & UX Consultant, and I took a quick look at their landing page. Here are 3 quick tweaks that can help drive even more local installs and conversions:

1. Instant Performance Hook: Highlighting exact speed improvements right in the hero section.

2. Frictionless Download Path: Making the primary action for desktop installation more prominent.

3. Clear Developer Quick-Start: Ensuring setup steps are instantly visible above the fold.

Would love to share the full detailed layout if the team is interested. Keep up the great hunting!

Best,

Abdul Moeez

AI CRO & UX Consultant"

@zaczuo Oh thanks for sharing! And appreciate all the support! We'll keep making Unsloth better and better for everyone!

A few folks asked the same great question (thank you for the great question!), so answering once here: how is this different from Notion/ClickUp/Drive with MCP?

Different layer. Notion is where your team writes docs; BearDrive is the folder your agents already work in, made shared. Our agents' output is mostly not prose (HTML reports, CSVs, decks), and moving those into a workspace is a step a human or an MCP call has to perform. That step is what fails, which is how reports end up as Slack uploads named report-v2-FINAL.html.

With BearDrive there is no step. The file syncs the second the agent writes it, with authorship down to the agent session, and a teammate's agent reads it at a real local path with its normal tools. Most teams keep Notion for docs and run BearDrive underneath for the working files.

@snowlee Nice launch congrats🙌i really like the focus on actual files rather than creating another separate AI workspace. It keeps the workflow simple.

@shuwn_jung let's goooooooo

@jay_janarthanan1 Not missing anything, it's a fair question (a few folks asked, so the short version):

Notion is where your team writes docs. BearDrive is the folder where your agents already work, made shared. Different layer.

Concretely: our agents produce HTML reports, CSVs, and decks on local disk. Moving those into Notion is a manual step (or an MCP call that turns a file into a page, lossy for anything that isn't prose). With BearDrive there's no step: the file syncs the second it's written, with authorship per change, and a teammate's agent reads it at a real local path with its normal tools.

So most teams keep Notion for docs and use BearDrive for the working files underneath. If your agents' output is mostly prose pages that end up in Notion anyway, your current setup is honestly fine. The difference shows up the day it isn't just prose.

@jaehee_ko Thank you very much for your support, Jaehee!

@michael_vavilov It's a fair question! Thank you, Michael. Honest answer first: if Notion + MCP is working for your team, don't migrate. Notion is great for prose the team reads.

The category is different though: we're much closer to Google Drive than to Notion. Notion is a place agents write into; BearDrive is a folder agents work in. And being the native file system makes a bigger difference than it sounds.

Our agents' output is mostly not prose: HTML reports, CSVs, PDFs, deck files. In Notion those become attachments, and an attachment is a dead end for the next agent. In a real folder they're just files at real paths, so when my agent continues a teammate's work it greps, edits, and re-renders with its normal tools. No fetching blocks over an API, no markdown-to-blocks conversion loss, no re-explaining context.

So in practice teams run both: Notion stays the wiki humans read, the shared folder is where working artifacts live, with attribution per change down to the agent session (which already caught an accidental rollback of a teammate's edits for us).

If your agents' output is mostly prose pages, you're honestly not our user yet, and that's fine. Curious though: what happens today when an agent produces something that isn't a page, like a CSV or an HTML report?

@galdayan Thank you for this question, Gal! It's the sharpest one here, and it's exactly the right thing to press on.

To your last point first: yes, there's a `.bdriveignore`, exactly gitignore-style, and it syncs with the folder itself, so the exclusion rules are team-wide policy rather than per-person discipline. `.env`, key files, raw scrape dumps, whatever patterns you define never leave the machine.

But it's not the only gate, and honestly not the main one. The main one is scope: BearDrive doesn't sync your whole project. You connect a specific subfolder (ours is `shared/`) and only what's written inside it ever syncs. Your agent keeps working in the project root as usual; scratch work and intermediate junk stay local. Writing a file into the synced subfolder is the send, the same deliberate act as posting to Slack.

And you control that behavior directly: the rules for what belongs in the shared folder live in your own AGENTS/CLAUDE.md ("deliverables only, never secrets or raw scrapes"), so it's an instruction your agents actually follow, versioned and team-visible like everything else.

When something wrong lands anyway: every change hits the feed with author, device, and agent-session attribution, files can be removed, versions rolled back. Recoverable in a way a Slack upload isn't. Though I won't oversell it: if a real API key syncs, rotate it. Same rule as git.

You're right that there's no review-before-sync step today. We're going to add more gates there, and if you have opinions on what it should look like (per-folder rules? a PR-style diff before something becomes the team's copy?), I'd genuinely love to hear them.

@samdcbu Thank you, and honest respect for the rclone + jujutsu rig, that's exactly the duct tape we keep finding in the wild, and "sync with provenance" is precisely the job. If BearDrive gets to retire it, I'll consider that the best compliment we've gotten.

On large files, to be transparent: BearDrive isn't specialized for synchronizing big files, but we've confirmed it handles 100 MB+ files without problems, so your `.lbdb` files should sync fine today. The one thing worth knowing is that history is kept per change, so a large file that gets rewritten very frequently costs storage and bandwidth on every write. If your agents touch the `.lbdb` constantly, you may want it in `.bdriveignore` with periodic snapshots synced instead; if it changes a few times a day, just sync it and don't think about it.

Your use case is exactly the kind that should shape how we handle big binaries going forward, so we'll take extra care there. Thank you for the kind words!

@fabcairo Thank you for your support, Fabio! Thank you! And honestly, it doesn't replace Notion. For docs humans write and collaborate on, Notion is the right choice.

What changed for us: once the whole team ran agents daily, most new artifacts weren't docs anymore, they were files on local disk: HTML reports, CSVs, decks (in HTML). Getting those into Notion or a Shared Drive takes a step someone has to perform, and that step is what fails. That's how reports end up as Slack uploads.

BearDrive removes the step. It syncs the folder agents already work in, so files are born shared: authorship down to the agent session, and teammates' agents read them at a local path with their normal tools, no API fetch, no paste.

So: Notion for the docs, BearDrive for the working files underneath. Curious, at your previous company, when an agent produced a CSV or an HTML report, where did it end up?

@omarships Congrats on the launch..🙌 using a sleek, swipeable UI to solve top-of-funnel list building is a super refreshing take on prospecting. Is there a cap on how many cards a user can swipe or export in a single session on the free tool?

@omarships I kept 11, but don't see the option to export, what's next?

@omarships Congratulations on the launch! The swipe mechanic makes this genuinely fun to use, but good prospecting usually comes from spending more time on fewer names, not less time on more. Has the format led people toward worse-fit lists because it encourages fast yes/no calls on thin information, or does the ranking do enough filtering upfront that speed doesn't actually cost accuracy?

@bengeekly didnt think of that, coming up next

@nikitaeverywhere Appreciate it

@kate_ramakaieva tinder for sales katie, exactly

@adiasg Built-in web search missing professional and marketplace data is a real gap. Which connectors are pulling live versus cached, and does the model know which one it got?

@adiasg Congratulations on the launch! Exciting to have followed this journey!

@yahia_bakour3 Thanks Yahia!

@sahil_darji1 This has been an incredible journey to watch from the growth side. 🚀

Over the past year, I’ve seen hundreds of developers build real, end-to-end production-ready applications with LaraCopilot. Seeing them go from an idea to a full-stack app with a real backend, database, auth, APIs, and code they actually own is still mind-blowing to me.

V2 is a massive leap forward. We’ve built something genuinely different from the typical AI app builders out there.

Proud of what the team has built. Now I can’t wait to see what people build with V2. 🔥

@rohanparmar This is exactly why we built LaraCopilot in the first place. ❤️

Seeing people turn an idea into something real, and then actually get users and revenue from it, has been one of the most rewarding parts of this journey.

Thanks for all the work you’ve put into making this happen, Rohan. 🚀

@dhavaljoshi007 Love this, Dhaval. 🙌

“your code belongs to you” part is something we’ve always cared deeply about. We wanted to build an AI app builder that doesn’t leave you with a prototype you have to throw away later.

Build it, ship it, own it. 🚀

@hareesh_vemasani Absolutely! That prototype-to-production gap is exactly what we’re trying to solve with LaraCopilot. V2 is going to push this even further more reliable builds, better agent workflows, and production-ready code while keeping developers in control. 🚀

@hareesh_vemasani Exactly! 🙌 The prototype-to-production gap is what we’ve been trying to solve from day one.

Really appreciate you calling that out, Hareesh. Excited for you to see what we’ve done with V2! 🚀

@alpeshnakrani This means a lot, Alpesh! 🙌

We’ve put a huge amount of effort into making the generated Laravel code actually useful and production-ready. Really glad you’re seeing the quality!

@mad94 Thanks Dorina! 🙌

Genuinely like Lovable, and honestly a lot of our users use both. The difference is what happens after the prototype.

Lovable generates your frontend and wires it to managed backend services. Fast for demos, but the backend is rented: your data, auth, and logic live on services you don't control.

LaraCopilot generates the entire stack as real code. Database schema, auth flows, background jobs, business logic, APIs. All of it in your repo. Export to GitHub, deploy to the cloud or your own server, hand it to any developer to extend. If you stop paying us, your app keeps running.

In short, Lovable is great for validating an idea fast. LaraCopilot is for when that idea needs to become a real product you own.

Happy to answer anything deeper on how the generation works!

@mad94 The biggest difference is that Lovable and most AI builders are stack-agnostic. LaraCopilot is Laravel-native.

@avinashvagh1 Couldn’t agree more, Avinash. 🙌

What’s been most exciting for us is seeing people go beyond prototypes and actually build products they use, ship, and grow with LaraCopilot.

V2 is the result of a lot of that feedback. Thanks for being part of the journey. 🚀

@akshay_shah10 Thank you!

🙌 That’s exactly what we wanted to achieve making it easier to go from an idea to a working Laravel app without leaving the Laravel ecosystem. 🚀

@akshay_shah10 Really glad to hear that, Akshay! 🙌

Keeping developers inside the Laravel ecosystem while making it dramatically faster to turn ideas into working products has always been a big part of what we’re building.

Thanks for using LaraCopilot! 🚀

@shruti_vaghasiya Thank you, Shruti! 🙌

That’s exactly what we wanted LaraCopilot to do: help developers go from idea to working software much faster.

Excited to have you all see what V2 can do. 🚀

@parikshit_builds For people like me with 3–5 cards across different banks, what’s the one “aha” moment you’ve seen where users suddenly realize, “Oh, this actually saves me real money without extra work”? And how fast does that usually happen after install?

@parikshit_builds The "no invented you-saved-$X math, only the banks' own stated ceilings" principle is something we've had to get disciplined about too, in a different corner of the same problem — bill negotiation and subscription savings are so easy to oversell with a big round number that isn't really earned. Sticking to numbers you can actually defend is a harder design constraint than it sounds, especially under pressure to make screenshots pop. Good to see it held as a hard rule instead of a nice-to-have.

@zara_noelle Honest answer, with one flat cashback card there's not much to optimize, the which card part earns its keep as your wallet grows. But the offers side works from card one. I just checked mine to put a real number on it, my Amex Blue Cash Everyday alone is holding 355 offers behind those add buttons, 1,469 across the wallet. That's real money at any wallet size. Same with the fee and credit reminders, one annual fee card is enough to forget one.

@landon_matthew Same, that screen is where I closed every other app too. Handing bank credentials to a third party just to see my own offers never sat right with me. That's basically why I built this the way I did. The extension just reads the page you're already logged into, nothing to hand over. Which app made you quit at that screen?

@gideon_henry Honestly yes. So the app shows its work. Every answer opens into the actual math, only the banks' own numbers, and it warns you when a bonus won't post, like Walmart counting as a superstore. The data behind it gets cross checked weekly and only publishes when sources agree, otherwise the app falls back to the card's base rate. And there's less to trust us with in the first place, no bank login, no Plaid. People trust the tool that admits when it doesn't know.

@alice_hayes2 Yes, exactly the habit it's built to break. Nobody loses money because they don't know their cards, it's because checking at the register feels like homework. So the answer has to be instant. Search the store, or just ask from your iPhone's search bar. Which card had become your everything card?

@sebastian_patterson Yeah, that moment is why I built this. Most offers only count if you added them before the purchase. So the RightCard extension just adds all of them in one pass from Safari, and by the time you're paying it's already on the card. Which bank's offers were you losing? Mine were Chase and Amex.

@iamtoruk this is actually so necessary

@iamtoruk Reading the session files the tools already write, instead of asking me to wire up another account, is the right design. Does the Optimize tab show what it changed before applying, or does it fix cache bloat automatically?

@adityavsingh Honest answer: I can see what a session cost. I have no way to see what it bought.
Most of my spend is not in the PR. It's the hour before it, where the agent reads the wrong files, I stop it, rewrite the prompt, start over. On a dashboard that's retry tax. But a good chunk of it is me working out what I actually want, and then the PR takes ten minutes. So cost per merged PR would make the messy session look awful and the clean one look great, when the messy one is what paid for it.
The metric I'd act on is one where the throwaway sessions get attributed to whatever they eventually produced. Right now I look at the weekly total and shrug.
Separate question: if one session touches three repos, where does the cost land?

@galdayan really good question, and its the exact case the attribution was built for, so let me be straight about where it works and where it doesnt.

Yes, there is a parent-child concept. when your orchestrator spawns subagents through Claude Code's Task tool, CodeBurn captures each spawn and ties it back to the PR the parent was working on when it fanned out. The side agents' spend folds into that PR, and the parent can be the anchor even if it barely spent anything itself. So your cheap synthesis session isnt "the one that touched the PR" while the real cost floats away unattributed. The PR row shows the parent plus all the folded workers, with a count of how many agent runs rolled in.

Two honest limits though:

it keys on native Task-tool spawning. If your fan-out is a custom harness or separate processes or cross-provider workers, that parent-child link isnt in any single tool's logs, so those fall back to weaker time/directory correlation or land in unattributed.

And it attributes to the PR that was active when the child was spawned. If the workers all run before the PR even exists and it only appears at the merge step, CodeBurn wont retroactively credit that PR with them. It leaves them unattributed rather than pretending the synthesis step "did" work that happened before it. Thats the same principle you noticed.

So the clean case (Task subagents on a branch you cut up front) folds properly. The messy case (the value only becomes a PR after the workers finish, or a mixed orchestrator) is exactly the frontier we are still working on, because that grouping only exists in your orchestrator's head, not in the logs. Genuinely the most interesting problem in this whole thing.

@gregberge Smart solution pinned feedback on exact image locations gives agents something much more useful than vague review comments

@gregberge Great idea giving coding agents a way to share screenshots directly in PRs could make reviews much clearer

@mateuszkonik nice to see you like it! On my side it runs Playwright (that is installed locally in the project for E2E + Visual Testing) and it browse the app locally, take screenshots or videos of the relevant parts and upload them. So it's not random, the agent takes care of using the good state. Also it's one-shot so we don't need the stability required for visual testing.

@samejr Hey James,

I came across Trigger.dev on Product Hunt today—huge congrats on the launch!

I'm an AI CRO & UX Consultant, and I took a quick look at your landing page. I put together a mini optimization blueprint that can help you reduce friction and boost your sign-up conversions:

1. Sharpening the Hero Value Prop: Highlighting immediate time-savings alongside technical capability.

2. Amplifying Social Proof: Moving developer trust badges higher up to drive immediate confidence.

3. Action-Driven CTAs: Swapping generic buttons for outcome-based prompts. Would love to share the full breakdown if you're interested. Keep crushing it!

Best,

Abdul Moeez

AI CRO & UX Consultant"

@yahia_bakour3 Thank you Yahia!

@abdullah_farouk This month we moved our entire company to Mac from Windows. I am looking forward to your team feature. Want to find out where people are wasting time

@abdullah_javaid3 Currently, each agent session is separated by design- so that there is no cross contamination and context leakage between the sessions. But, in future, we may introduce a custom header where user can define a Cohesor sessionID so that few agent sessions can be combined together in the gateway.

@busmark_w_nika coming soon

@statewavedev  @saber_maram Outstanding launch, Building a governance-first memory runtime addresses the exact missing layer in modern agent architecture.

@statewavedev  @saber_maram Congrats on the launch, Saber! Not a pain I have hit yet, but the one I expect first is what happens when many agents share the same memory at once. Every time an agent asks for something, the system has to decide right then what that agent is allowed to see. That is hard to add later, which is why designing for governance from day one is the right call.

@statewavedev  @saber_maram I’ve been following this project for a while, and this feels like a real step toward making agent memory auditable, controllable, and dependable. I’m especially excited to see how the open-source community contributes to it.

Congratulations, and good luck!

@istiakahmad Congrats-back 🙏 Yes — that's exactly what the admin console is for. A subject explorer lets you browse and drill into per-subject memories and episodes, inspect provenance, state-assembly receipts, sensitivity-label policy and compile-job health — and manage compiled memories directly (edit/invalidate), no raw SQL against Postgres.

If your operators prefer the terminal, there's a CLI with full parity — every operation the web UI does. It's a privileged, secure-by-default console (password gate + access gateway), meant to run against your own instance.

This is the repo: https://github.com/smaramwbc/statewave-admin

What's your operator workflow — mostly audit/inspect, or hands-on correction?

@rajesh_gupta__ Facing the same problems is why we build this. Share your experience we are open for all response.

@abhi_singh4000 Thanks - that's where we build the https://github.com/smaramwbc/statewave-multi-agent-memory for. Allways open for any comments on it.

@arun_pratap_singh7 Thanks for your comment, let me know when you face questions.

@arvish_suresh Great question. Short answer: self-hosted + open source stays the core — that's the whole point (own your data, inspect the governance yourself, no phone-home), and it's not going anywhere.

That said, yes — a managed/hosted option is on the roadmap for later, for teams who'd rather not run the infra themselves. The important part: the OSS won't turn into a crippled "community edition" to push people to cloud. Self-hosted stays fully supported and first-class.

Out of curiosity — is it mainly the infra/ops overhead you'd want to skip, or something else? That helps us shape what the hosted version should prioritize.