AI Product Updates, August 2: Five Launches for Personal Reps, Offline AI, and Agent Workspaces
Five August 2 launches gave AI products a phone identity, local file access, recoverable agent sessions, on-device speech, and India-first voice infrastructure, while no same-day frontier-model or API change was confirmed.
Five launches on August 2 pushed AI products out of the chat window: Zinley gave an agent its own phone number and inbox, LumiChats and Termexo put agent work on the desktop, Zen Whisper kept speech recognition on the Mac, and Bolcho packaged India-first voice infrastructure. The official product indexes checked for OpenAI, Google Gemini, and GitHub showed no same-day frontier-model release, API price or quota change, or deprecation notice. 123
This issue covers product releases and announcements dated August 2, 2026 in the UTC+8 publishing window. The five launch entries below come from the visible Product Hunt daily board and their detail pages or official product sites. Product Hunt's all-products view returned an access challenge, so this is a confirmed sweep of the accessible daily launch surface, not a claim that every small launch was captured. 4
At a glance
| Product | Releasing entity | Release date | What changed | Availability |
|---|---|---|---|---|
| Zinley | Zinley team | August 2 | A personal agent with its own phone number, inbox, and computer; it handles calls, email, and computer work within user rules 5 | Start free; Product Hunt lists free options 6 |
| LumiChats Free / LumiChats Offline | LumiChats team | August 2 | Moves the product from browser chat toward a desktop that can touch files and run local workflows; the product page also describes offline model use and local documents 7 | Free options listed; open-source offline variant for Windows, Linux, and macOS 7 |
| Termexo | Termexo team | August 2 | A local Windows workbench for Claude Code and Codex with session resume, approval notifications, and model profiles 8 | Free; Windows 10/11; no account or server 9 |
| Zen Whisper | Zen Products | August 2 | Cursor-following Mac dictation, file and media transcription, searchable history, and 110 spoken-language options 10 | Available now; 14-day trial, 3,000 words/week free tier, and paid seat packs on Apple Silicon macOS 14+ 11 |
| Bolcho AI | Bolcho team | August 2 listing | India-first voice and chat agents for phone calls and websites, with inbound/outbound calling, a web widget, and an API 12 | Free to start; the site says no credit card is required; production pricing is not disclosed 412 |
Agents get a number and a workflow
Zinley gives the agent an identity
Zinley is built around a different boundary from a chat assistant. The product says the agent has its own phone number, inbox, and computer, remembers the user's people and open loops, and acts within rules set by the user. It can take or make calls, handle email, book meetings, and perform work on approved devices. 5
The availability boundary is simple: the official site says to start free, while Product Hunt marks the launch with free options. The launch page does not provide a rate card or explain how usage is metered. 56
The useful test is whether Zinley's identity model creates a better handoff or a larger blast radius. A phone number and inbox let the agent remain in a thread, but they also make mistakes look like they came from the user. Before connecting real contacts, test approval points, call transcripts, calendar actions, and the exact device permissions that can be granted. The official site lists integrations including Gmail, Slack, Google Calendar, Notion, Linear, GitHub, Drive, Figma, Stripe, iMessage, and Discord; that is a product capability list, not an independent security assessment. 5
Bolcho targets phone work in India
Bolcho AI is an India-first platform for building and deploying voice and chat agents for phone calls and websites. Its official site lists inbound and outbound calls, an embeddable web widget, lead capture, conversation analytics, knowledge-grounded answers, CRM integrations, and an API with webhooks and scoped keys. It says the platform supports 11 or more Indian languages, including Hindi, Hinglish, Tamil, Telugu, and Bengali. 12
This is a production layer rather than a single voice model. The site says teams can switch among providers for speech recognition, language models, text-to-speech, and telephony, and can bring their own keys or SIP trunk. It also advertises near-provider-cost pricing and no markup on minutes, but no public plan or rate card was returned, so those claims cannot yet support a cost comparison. 12
Product Hunt placed Bolcho on its August 2 daily board; the official site does not show a separate dated launch announcement. The current availability claim is free to start with no credit card required. For a real deployment, the missing decision variables are call recording and retention, language-level quality, provider failover behavior, and the total cost after telephony and model usage. 412
Local desktops become the control plane
LumiChats moves beyond the browser
The August 2 LumiChats launch is presented as a desktop counterpart to the browser product. Its Product Hunt page says the desktop app can touch the user's files, run commands on the machine, and write real documents. The same page groups this launch under LumiChats Offline, a free open-source app built on GPT4All that supports Mistral, Llama, Qwen, DeepSeek, and LumiChats models, plus local document retrieval. 7
The previous state matters here: browser chat is convenient, but it cannot directly operate on local files. The desktop boundary removes that limitation while adding a new one: the user must understand which commands the app can run and which files its local-document feature can read. Product Hunt lists free options, but the launch copy says users pay for the work they run and gives no public usage rate. 7
The first trial should use a throwaway folder containing a few PDFs and a reversible document task. Check whether local mode really keeps prompts, source files, and responses on the machine, then inspect the permission prompts before pointing it at a work repository. The page's privacy and model-support statements are product claims, not an audit. 7
Termexo makes agent state visible
Termexo packages Claude Code and Codex into one recoverable Windows workspace. Its launch page lists real PTY terminals in configurable grids, searchable and resumable native sessions, notifications when an agent needs approval, and Claude-compatible model profiles. 8
The official site adds the operational boundary: Termexo runs on Windows 10 or 11, needs no account or server, stores API keys through Windows Credential Manager, and reads agent session files without rewriting or deleting them. It is available as a free local app through an installer or
npx termexo@latest. 9Termexo's important change is not another model selector. It is the control surface around several agents: one panel shows which task is running, waiting for approval, thinking, or finished, and one click resumes the matching session. The product site labels phone approval and quota visibility as planned rather than shipped, so those should not be treated as current capabilities. 9
Speech keeps its local boundary
Zen Whisper turns the cursor into a microphone
Zen Whisper is a native Mac dictation and transcription app from Zen Products. Hold a shortcut and it inserts speech into the active text field; it can also transcribe uploaded audio, video, and supported public links, then keep the results in searchable history. The maker says core speech recognition runs on the Mac and exposes 110 spoken-language options. 1011
The availability is more detailed than the Product Hunt label suggests: the official site says the app is available now for Apple Silicon Macs on macOS 14 or later, starts with a 14-day free trial, remains usable for 3,000 words per week on the free tier, and sells paid seat packs with one year of updates. It also warns that optional network-backed features are separate from the core local recognition path. 11
That split is the feature to verify, not the language count by itself. Test dictation with the network disabled, then test media transcription and translation separately. A local speech pack can reduce exposure of microphone audio, but it does not automatically make every optional workflow local. 11
What changed in practice
August 2 produced no new frontier model or API pricing move in the checked official release surfaces. The new products instead filled in the layers around existing models: Zinley supplies identity and communication channels, Bolcho supplies telephony and provider routing, LumiChats supplies file access, Termexo supplies supervision across local agent sessions, and Zen Whisper supplies a local speech input path. 1235791112
The common product question is becoming more concrete: where does the agent act, what does it get permission to touch, and how can the user recover the work when it stops? These five launches answer different parts of that question. They also make the trust boundary easier to inspect because each product claims a specific place for identity, files, sessions, audio, or provider routing.
What to test next
- Identity and delegation: Put Zinley on a test phone number and mailbox. Measure when it asks for approval, what it reports after a call or email, and whether a user can revoke device and integration access without deleting the work history. 5
- Local file access: Use LumiChats with a throwaway folder and a small document set. Compare the offline path with the desktop workflow, record every command or file write, and verify whether any optional model or service sends data outside the machine. 7
- Agent supervision: Run two non-destructive Claude Code or Codex tasks in Termexo. Check whether the approval banner identifies the correct project, whether a closed session resumes with full context, and where keys and session files are stored. 89
- Speech privacy: Test Zen Whisper with network access disabled, then enable only the features that need a network. Compare local dictation, file transcription, and translation as separate data flows rather than treating the app as uniformly offline. 11
- Voice-agent economics: For Bolcho, run the same short call through two provider configurations and log telephony, model, speech, and retry costs separately. Verify language switching, failover, recording retention, and API-key scope before using its near-cost pricing claim in a budget. 12
References
- 1OpenAI product releases
openai.com
- 2Gemini API changelog
ai.google.dev
- 3GitHub Changelog
github.blog
- 4Product Hunt daily leaderboard, August 2, 2026
producthunt.com
- 5Zinley official site
zinley.com
- 6Zinley launch page
producthunt.com
- 7LumiChats launch page
producthunt.com
- 8Termexo launch page
producthunt.com
- 9Termexo official site
termexo.com
- 10Zen Whisper launch page
producthunt.com
- 11Zen Whisper official site
zenproducts.ai
- 12Bolcho official site
bolchoai.in
This story was produced automatically by a channel. One sentence is all it takes for Neodrop to keep producing for you.