Every feature, improvement, and fix we ship lands here first.
Analytics has a new section: Fixes you can apply. Press Scan and Granetra reads the questions your bot couldn't answer, finds the ones that keep coming back, and proposes something specific for each โ not "consider adding more docs". Every suggestion shows the exact questions behind it, and nothing at all changes until you press Apply.
The best suggestion it can make is one it didn't write. When your bot missed a question and a colleague later answered it by hand in a live chat, Granetra offers to reply with their exact words next time โ shown to you in full before you apply it, because it's going to customers unedited. Replies that mention one customer's order number, email or address are never offered: those belong to that conversation.
Where nobody has answered a repeated question yet, Granetra offers a short flow that admits it doesn't know, takes the visitor's email and fetches a human โ so they leave as a lead instead of a dead end. And if your website hasn't been re-read in a while, it offers to catch up with what you've published since, which is often the whole fix. Applying a flow is a Scale feature; re-reading your site is included from Growth.
Some gaps only you can close, and those are marked as exactly that โ no button, no invented answer. Granetra will not write your refund policy and offer to publish it. It names the questions, tells you how often they came up, and leaves the rest to you.
If your shop runs on WooCommerce rather than Shopify, it now connects the same way: Bot Studio โ Store & payments, paste your store address and a REST API key pair, and we check them against your live store before saving anything. Your bot then answers from real data โ where an order is, what's in it, whether something is in stock and what it costs, what's still refundable. Refunds, cancellations and delivery-address changes work too, and land in exactly the same approval queue as everything else. Included from Growth, like the other store connectors.
WooCommerce has no tracking field of its own โ every store gets one from a plugin, and they all store it differently. Your bot reads the common ones, so "where's my parcel?" gets a carrier and a tracking number where your store records them. Where it doesn't, the bot says so plainly instead of inventing a number, which is the failure that actually costs you a customer.
New in the sidebar: one screen that checks your team against the controls a security review asks about, and tells you which ones don't currently hold. Who signs in without a second factor, invitations left open for weeks, members who haven't logged in for months, API keys overdue for rotation. Each one says what it proves and exactly how to fix it. A check we couldn't run says so in its own colour โ it never quietly counts as a pass.
Auditors don't ask whether you have roles; they ask who last checked that everyone still needs theirs. Read through every person, invitation and API key with access to your team โ the ones worth a second look are sorted to the top โ then certify it in one click, with a note on anything you decided to keep. The record is permanent and can't be edited or withdrawn afterwards, which is the only reason it's worth anything. Reviews are due every 90 days and the console tells you when the next one is.
Team owners can now make a second factor mandatory. A member who hasn't set one up keeps their own account and their own bots, and simply gets nothing through the team until they enrol. It never applies to someone opening a bot they built themselves โ a security setting shouldn't be able to lock a colleague out of their own work.
Every threat Granetra watches for now has a page: what it means, how to tell a real one from a false alarm before you act, what the system already did on its own, and the steps after that. It's generated from the same detector settings the live system uses, so it can't drift into telling you something that isn't true while you're in the middle of an incident.
Download your team's full change history as a CSV โ who did what, when, to which thing. Handy for a security questionnaire, a customer's due diligence, or your own records. Alongside it, a published list of every service we rely on and exactly what each one can see.
Our own internal accounts โ the ones that can reach any workspace for support โ can no longer be held with a password alone. This is about our side of the trust, not yours, and it's the control most often asked about in an enterprise security review.
Change history could previously have been edited or deleted by the application. It can't be any more โ the database itself refuses, so "nobody rewrites the record" is enforced rather than promised. Nothing in Granetra was ever doing so; the difference is that now nothing could.
Connect Zendesk in Bot Studio โ Helpdesk and the two systems become one. A chat that needs a person opens a ticket with the transcript attached โ exactly one, even if the visitor asks twice. A reply typed in Zendesk appears in the customer's chat within seconds; a reply typed in Granetra appears on the ticket. Resolve it in either place and it resolves in both, and the conversation hands itself back to the bot. Every conversation in your inbox shows where it lives, with a link straight to the ticket.
When a ticket is opened or a customer writes in, the bot writes a suggested answer from your knowledge base and posts it as a private note. Your agent reads it, edits what they want and sends. Nothing drafted ever reaches a customer on its own โ the note is internal, every time, and you can switch it off per bot.
Import your published Zendesk Guide articles and the bot answers from the documentation your support team already maintains โ including the articles behind a login that a website crawl can never reach. They appear alongside your other knowledge sources, and removing them is one click. Drafts are skipped, because an unpublished article is usually unpublished for a reason.
The bot can look a ticket up, file one, reply on it, escalate it or mark it solved. Looking one up is free; every action that changes a ticket waits for a person by default, lands in Approvals with exactly what was asked for, and only runs when someone releases it. On Growth that's permanent. Scale lets you release the ones you trust โ and every ticket action shows up in the Permissions console with everything else your AI can do.
Webhooks get lost โ providers drop them, deploys race them. If a reply from Zendesk hasn't arrived, the chat window catches it on its next check rather than leaving someone staring at an answer that was already written, and a scheduled pass repairs anything still adrift. A retried delivery can never show the same message twice.
Your bot can work your store, not just talk about it. In Bot Studio โ Store & payments, paste a Shopify Admin API token and a Stripe restricted key (we check both against the real thing before saving) and your assistant starts answering from live data: where an order is, who's carrying the parcel and its tracking link, whether something is in stock and what it costs, what a customer paid, what plan they're on and when it renews. Nothing to host, no endpoint to build, no JSON schemas to write โ the actions are built in. Store connectors used to be Scale-only; they're now included from Growth.
The bot can also take the requests that cost you money: issue a refund through Shopify or Stripe, cancel an order, fix a delivery address before the parcel leaves, or cancel a subscription at the end of the period the customer already paid for. Every one of those is held for a person by default. The customer is told the request is with your team, it lands in Approvals with exactly what was asked for, and nothing happens until someone releases it. On Growth that's permanent โ anything that moves money always waits for you.
Scale adds the policy layer: decide per action what runs on its own and what comes to you, including a cash limit. Set refunds to โunder 25โ and a small goodwill refund goes through mid-conversation while anything larger waits for a person โ and a request with no amount on it, which means the whole remaining balance, always waits, whatever the limit says. The approval that lands in your queue tells you which rule sent it there.
If a Scale account moves down to Growth, any store action that was set to run unattended goes straight back behind the approval queue โ immediately, without anyone having to remember. Read-only answers like order status and tracking keep working exactly as before, so a billing change never leaves a customer without an answer. Turning a connector off, or putting an action back behind a human, works on every plan and always will.
Point a โSend Messageโ ad at your connected WhatsApp number and your bot opens the chat already knowing the ad's headline โ so it answers about that product or course straight away instead of asking someone who just told you what they want to repeat themselves. Nothing to configure. Each conversation is tagged with its ad in your inbox, and the new Analytics โ By ad panel ranks campaigns by conversations, engagement, emails captured and answer rate โ showing you when an ad is promising something your bot hasn't been trained on yet.
Leads delivered to your webhook or CRM now include the phone number, every field the conversation collected as structured data (not just a summary line), which channel it came from, and the ad that produced it โ including Meta's click id so you can report conversions back. Existing fields are unchanged, so integrations you've already built keep working exactly as before.
A workflow that captured a lead required an email, which meant an enquiry from WhatsApp โ where the customer's verified phone number is already known and many people never type an email โ could silently fail to reach your CRM. A phone number is now enough on its own.
The website now speaks Sinhala, Tamil, Hindi, Chinese, Spanish, French, Arabic, German, Japanese and Korean alongside English. Pick your language from the new selector in the navigation bar, or just arrive โ first-time visitors are greeted in their browser's language automatically, and your choice is remembered. Each language lives at its own address (granetra.com/si, /ta, โฆ), Arabic lays out fully right-to-left, and the chatbot has always replied in whatever language you type.
The new AI Employees page shows every specialist you've built as a member of staff: what it's there to achieve, which tools it can use, how many conversations it has handled, how many it resolved, and the leads it brought in. Growth includes one AI employee; Scale lets you build a team of up to eight who hand work to each other. Find it in the dashboard sidebar.
You can now mark any agentic action as needing a person's approval โ refunds, discounts, sending a contract, deleting a record. The AI still gathers everything and asks for it, but instead of running the action it files a request and tells the visitor it's under review. Nothing happens until you approve it in AI Employees โ Approvals, where you see exactly what was asked, by which employee, and what the visitor said. Requests expire after 24 hours if nobody answers.
Each specialist can now carry a goal โ "get the visitor booked with the fewest questions possible", "resolve issues on the first reply" โ separate from the topics it handles and the way it behaves. It works toward that goal while it helps, and is explicitly never allowed to pressure a visitor or overstate what you offer to get there.
Scheduled re-syncing โ the bot re-reading your website on its own every day, week or month so its answers never go stale โ is now a Scale plan feature, and it's listed on the pricing cards. Re-syncing by hand from Bot Studio โ Re-sync website is unchanged and still available on every plan, as often as you like.
Every plan now states its AI knowledge capacity up front: how many pages of your website the bot reads on each sync (500 on Starter, 2,000 on Growth, 10,000 on Scale), how many documents it can hold (20 / 50 / 200 per bot), and the largest file you can upload (25 MB, or 50 MB on Scale). No more guessing whether your site or your document library will fit. Enterprise is custom โ talk to us.
Website training and re-syncs now read as far as your plan allows instead of stopping at a single flat ceiling, so larger sites get learned more completely. Deleting a document in Manage knowledge frees its slot straight away.
Enterprise moved out of a footnote below the pricing grid and into a proper fourth card, so all four plans compare side by side on the homepage and in your dashboard.
Describe when something should happen ("when someone wants to book a demo") and what you need to know, and your bot takes it from there: it recognises the request however the visitor phrases it, asks for whatever is still missing โ never twice for something they already told you โ checks that emails and phone numbers are actually valid, and only then emails your team, saves the lead to your CRM, calls your own systems, or brings in a real person. It will never tell someone their booking is confirmed unless it genuinely went through. Start from a Book-a-demo, Capture-an-enquiry or Support-request template in Bot Studio โ Automation โ AI workflows.
On WhatsApp, Telegram and Slack, customers send pictures and voice messages constantly โ a photo of a receipt, a screenshot of an error, a quick recorded question. Until now your bot ignored those completely and the customer got no reply at all. Now it reads the image (including any text, order numbers or amounts in it) and transcribes voice notes, then answers as normal. Files it can't open are acknowledged politely and the customer is asked for the details instead of being left in silence.
If a question lands with the wrong specialist on your AI team, they now pass it straight to the right one and the customer gets the correct answer immediately โ no more waiting for the next message to be re-routed. The handover is silent, and the receiving specialist answers under their own instructions, tools and model. Your inbox now also labels every reply with the specialist who actually wrote it, so you can see exactly how a conversation moved through your team.
When visitor memory is on, the chat window now carries a quiet โForget meโ link that erases everything that bot has stored about that person, instantly โ no account and no support request needed. It stays visible on white-label widgets too. One less privacy question to answer in a procurement review.
Bots have always answered in the visitor's own language. You can now choose the script they write it in: proper native script (เทเทเถเทเถฝ, เฎคเฎฎเฎฟเฎดเฏ, ุงูุนุฑุจูุฉ), matching however the visitor typed, or romanized Latin letters. Native script stays the default โ most people read their own script even when they type phonetically.
Settings save themselves the moment you change them, which left nothing to cancel if you changed the wrong thing. Editing surfaces now show when the last save happened and offer an Undo that steps back through your recent changes one at a time.
Per-agent tool permissions now list your MCP connectors alongside your actions, so you can give one specialist access to a connected system without giving it to the whole team. Granting a connector grants every tool it exposes โ MCP servers add tools over time, so permission is per connector rather than per tool.
โTalk to a humanโ moves the conversation to a dedicated full-page chat instead of the small widget panel โ roomier bubbles, live status, and the transcript so far carried straight across. Visitors already signed in to Granetra go right in without being asked to log in again; only anonymous visitors see a sign-in step, and it drops them back into the same chat. Replies still show up in the widget too, so they can switch tabs freely.
Analytics now shows how many people are visiting your website, not just how many chat with your bot. Because your widget is already on every page, there's nothing new to install โ traffic starts appearing automatically. You get visitors, pageviews, sessions and pages-per-visit, each compared against the previous period, plus a timeline you can filter to daily, weekly, monthly or yearly, and breakdowns of your top pages, where your traffic comes from, and what devices people use. A live counter shows how many visitors are on your site right this minute, and the whole panel refreshes on its own while you watch. Query strings are never recorded and no advertising cookies are set.
Granetra now watches its own front doors continuously. Sign-in, the staff admin panel, the public API, your chat widget, and the website crawler all feed a detection engine that flags unusual behavior โ password-guessing runs, credential stuffing, API keys used from addresses their owner never allowlisted, prompt-injection attempts against your bot, and requests aimed at internal addresses. Individually harmless events are correlated too, so a burst of small oddities from one source is recognized as the attack it is. Anything high-risk alerts our team instantly and, if nobody has acted on it within two minutes, the platform defends itself automatically: the source is blocked, its sessions are terminated, and further requests are refused. Every incident is kept with full forensic detail โ source address, device, timestamp, attack type, and exactly what was done about it, by a person or by the system.
Scale bots can now speak MCP (Model Context Protocol), the open standard for connecting AI to external tools. In Bot Studio โ MCP connectors, paste any MCP server's URL (plus an auth token if it needs one), hit Test connection, and the bot instantly discovers every tool that server exposes โ search a docs site, query an internal system, look up live data โ no per-endpoint registration like agentic actions. Discovered tools join the same tool loop the bot already uses: it decides mid-conversation when to call them and answers from the live result. Tools re-sync every time you save the connector, and each one can be paused with a switch.
Growth and Scale bots now go wherever your customers already are. In Bot Studio โ Channels, connect WhatsApp (via Meta's Cloud API), a Telegram bot, or a Slack app โ paste your platform credentials, follow the built-in numbered setup guide, and hit Verify connection to check your tokens against the platform's own API before going live. The same bot answers there with everything it has on your website: the full knowledge base, conversation flows, your AI agent team, and persistent visitor memory (a returning WhatsApp number is recognized as the same person across sessions). Every channel conversation lands in your one unified inbox with a channel badge, and when you join a conversation as a human, your replies are delivered straight back to the visitor on their own app.
Choosing a model for your bot (or for an agent in your AI team) no longer means typing a raw model id. Each provider now shows a curated dropdown of vetted models with friendly names and speed/cost hints โ GPT-5 Mini, Claude Sonnet 5, Gemini 2.5 Flash, and more. No more silent typos: previously a mistyped id would save fine and quietly run the default model instead. Bots already saved on a custom id keep working unchanged, and if you need a model that isn't listed, the Request-a-model flow is right below the picker.
Scale bots can now be a routed team, not a single assistant. In Bot Studio โ AI agent team, create named specialists โ Sales, Support, Billing, HR โ each with its own description, optional specialist instructions, its own model if you want one, and an explicit allowlist of which agentic actions it may call. Start in one click with the ready-made Sales ยท Support ยท Billing starter team (or pick a template when adding an agent) and tweak from there. A supervisor silently reads every visitor message and routes it to the right specialist (staying with them while the topic continues), specialists share the bot's knowledge and visitor memory, and the widget shows a subtle label over each reply naming who answered. Visitors just see one seamless assistant; tool permissions are enforced on every single message.
Growth and Scale bots can now remember people, not just conversations. Turn on Visitor memory in Bot Studio โ Behavior & messages and the bot quietly builds a lightweight profile of each visitor from what they share in chat โ their name, company, preferences, and what past conversations were about. When they come back days or weeks later (even in a new tab), the bot recognizes them, greets them like a known customer, and picks up context from previous sessions instead of starting from a blank slate. Leaving an email in the chat links it to the same profile automatically. Off by default, one switch to enable per bot.
Growth and Scale plans can now dress the chat widget in one of 12 professionally designed themes โ e-commerce, SaaS, healthcare, real estate, education, hospitality, finance, legal, fitness, agencies, travel, and a universal minimal look. Every theme ships with matching dark and light variants, and with the color mode set to Auto the widget follows your website's own light/dark toggle live. Pick one from Bot Studio's new Widget theme section (or right after training a new bot) and preview it instantly before saving. The classic Granetra look stays the default.
A complete guide to Granetra now lives at /docs โ training a bot, every Bot Studio panel (appearance, behavior, knowledge), automation (proactive triggers, agentic actions, conversation flows), conversations and analytics, teams and roles, a full plan comparison table, billing, security, the API, and a troubleshooting section for common issues. Linked from the nav bar and footer.
Every paid plan โ Starter, Growth, and Scale โ now shows a blue verified badge next to your account name in the dashboard, so it's clear at a glance that yours is a real, subscribed Granetra account. Granetra staff (Chief accounts) carry it by default. See it on your dashboard sidebar, and check the plan comparison table in the docs.
Build Granetra into your own stack. Scale teams get a new Developers page with API keys and webhooks. Create a scoped key and call the REST API to list your bots, read conversations, or send a chat message and get the bot's reply โ all from your own code. Point webhooks at your URLs to get a signed, retried POST the moment a conversation starts, a visitor asks for a human, or a lead is captured, with a full delivery log, one-click test events, and re-delivery for anything that failed. Every webhook is HMAC-signed so you can verify it really came from us. See the built-in API reference on the Developers page to get started.
Managing your team got a lot smoother. Pending invites now show a live countdown to expiry and can be re-sent with one click instead of revoking and starting over. When a team gets bigger you can search members by name or email, and select several at once to change their role or remove them in a single action. The invite box catches common email typos (like gmial.com) before an invite quietly goes nowhere, remembers the role you invite people as, and tells you clearly when your seats are full. The activity log can now be filtered by action or person so you can find exactly what changed.
When you invite someone to your team, they now get the full abilities of your plan across the whole dashboard โ not just on the surfaces we'd already wired up. A teammate on a Growth or Scale team sees your bots as Growth/Scale-capable in Bot Studio, can open advanced analytics, and can create bots in the team, even if they've never bought a plan themselves. Everything follows the team owner's plan, so an invited member is never wrongly shown locked features or an upgrade prompt for things your plan already covers.
Your account is now a workspace you can invite teammates into (Growth includes up to 3 seats, Scale up to 10). Head to the new Team page in your dashboard, send an email invite, and pick each person's role: Admins can create and delete bots and manage the team, Members can build and run bots, and Viewers get read-only access to bots, conversations, and analytics. Everything your team does โ bot changes, knowledge edits, live-chat takeovers, invites โ is recorded in a per-team activity log, and each teammate only ever sees the workspaces they belong to.
Your bot's answers in the widget now display Markdown formatting โ bold, italics, bullet and numbered lists, inline code, and links all render properly instead of showing raw ** and - characters. Formatting is sanitized safely, so nothing a model outputs can inject anything into your visitor's page.
Scale bots are no longer locked to a single vendor. In Bot Studio โ Behavior & messages, pick the AI model that answers for each bot from a redesigned model gallery โ Platform default (our tuned OpenAI setup), or bring your own via OpenAI, Anthropic (Claude), Groq, DeepSeek, xAI (Grok), or Gemini, with room to type any model id the provider offers. An unrecognized model quietly falls back to the default, so your widget never breaks.
Granetra now has an Enterprise tier, starting at $499/month: custom conversation volume sized to your traffic, dedicated onboarding where we build and train your agents with you, priority support on a direct channel, contract-based invoice billing, and a DPA for your legal team. It's sales-led โ visit the new Enterprise page and book a demo, and we'll come to the call with an agent already trained on your own website.
The gear button on a bot now opens the Bot Studio: a full command center with everything for that bot in one window, organized into clearly separated sections โ Appearance, Behavior & messages, Access & integrations, Knowledge sources, Add knowledge, Re-sync, Agentic actions, Proactive triggers, Conversation flows, the embed code, and a danger zone. No more hunting through a dropdown of separate pop-ups: click the gear once and switch between sections from the sidebar, with your bot's name and live status always in view.
Training a bot on a website now works on far more sites โ we use a browser-like crawler that isn't turned away by common bot-protection, and we retry once if a site hiccups. If a site still can't be read (it blocks crawlers, is down, or is built entirely in JavaScript), you now get a plain-English explanation and what to try instead, rather than a confusing failure. And if you train on a PDF plus a URL, a problem with one no longer throws away the other โ the bot is built from whatever we could read.
When one of your team is chatting with a visitor live, messages now appear noticeably faster on both sides. And if an agent steps away without handing the chat back, the visitor gets a clear "Back to bot" button in the chat header โ so they're never stranded waiting for a reply, and the assistant picks things back up automatically.
If our team has given you a plan for a fixed term, your dashboard now tells you the date it runs until โ right under your conversation usage. No more wondering how long you've got. It sits alongside your usual monthly reset date, which is unchanged: your conversation allowance still refreshes every month either way.
A visual refresh across every bot panel โ Configure, Agentic actions, Proactive triggers, Conversation flows, and Knowledge. The option cards (tone, theme, launcher position, and more) now sit in tidy equal-height rows with a clear selected-state dot, colour swatches and sliders got a crisper look, list rows lift gently on hover, and empty states feel more alive. Same settings, easier on the eyes.
Scale plans can now build guided, multi-step conversation flows that run before the AI. Pick trigger phrases (like โrefundโ or โbook a demoโ), then script the steps: say something, ask a question and save the answer, offer tappable choice buttons, branch on what the visitor says, or hand off to a human โ all from a visual builder on the bot card. Anything a flow doesn't cover still gets the normal AI answer, so flows add tight control exactly where you want it without giving up everything else.
Scale plans can now restyle the chat widget with their own CSS โ colors, borders, fonts, spacing, anything. Paste your stylesheet in the bot's configure panel (Appearance โ Custom CSS) and it's applied inside the widget only, safely isolated from the rest of your site, with your rules taking priority over the defaults. Perfect for matching the widget to your brand pixel-for-pixel.
Growth and Scale bots can now message visitors first instead of waiting to be asked. Set up to five rules per bot: reach out after someone's been on a page a while, catch them right as they're about to leave, or greet visitors on specific pages like /pricing. The message appears as a small chat bubble by the launcher โ clicking it opens the conversation with your opener already inside. Each visitor sees a given trigger at most once per visit, and only one nudge shows per page, so it never feels pushy.
A design pass over the dashboard, homepage, and sign-in pages. Dashboard stat cards now glow in their own accent color, the sidebar highlights your page with a gradient marker, usage meters shimmer, and empty states feel alive. On the homepage the hero headline animates through the brand colors, the most-popular plan stands proud of its neighbours, and testimonial and sign-in cards picked up finer gradient details.
The Settings page got a full visual refresh โ a sticky section navigator on the left (Account, Security, Preferences, Billing & data, Danger zone) that highlights where you are as you scroll and jumps to any section in one click, plus clearer section cards with icons for profile, email, password, two-factor auth, sessions, and more.
When a visitor asks to talk to a human, the widget now shows a clear waiting state โ a spinner, how long they've been waiting, and a Cancel button. If nobody on your team is available, the visitor can leave the queue and continue with the assistant instead of being stuck, and a gentle note after two minutes lets them know the team seems busy. Reloading the page while waiting keeps their place in the queue.
Turn on live chat and a "Talk to a human" button appears in your widget. When a visitor asks for a person you're emailed instantly and can jump into the conversation from your inbox โ your replies appear in their chat window in real time, and you can hand the chat back to the bot when you're done. The inbox flags every conversation that's live or waiting for a human.
Each bot's settings menu now has a "Reset analytics" action that clears its stored conversations โ wiping the analytics dashboard and conversations inbox back to zero for a fresh start, without touching the bot, its knowledge, or its settings.
Analytics now pairs conversation volume with an answered-rate trend on one timeline, groups top questions by topic instead of raw messages, shows where chats drop off, surfaces missed opportunities (repeated questions your bot couldn't answer), and suggests concrete next steps โ each KPI with a 30-day trend.
Connect your own API endpoints as actions your bot can take mid-conversation โ look up an order, check availability, create a ticket โ with a built-in endpoint tester and per-action on/off switch.
Set a bot trained on a website to automatically refresh its knowledge daily, weekly, or monthly, instead of only ever re-syncing by hand.
Chat widgets now offer 10 free messages per conversation before asking a visitor to leave their name and email to keep chatting โ a lighter-weight lead capture for longer conversations.
Hovering a point near the peak of your conversation volume chart now always shows its tooltip instead of getting clipped off-screen.
Configure, Re-sync website, Add knowledge, Manage knowledge, and Delete now live in a single menu on each bot card instead of five separate buttons.
See every PDF and website you've trained a bot on, with a chunk count and the date it was added, and remove any single source without touching the rest.
Upload up to 8 files at 25 MB each when training or adding knowledge, up from 5 files at 15 MB.
Set how wide the chat panel opens, alongside the existing height control, from a bot's settings.
Dedicated pages for every industry Granetra serves, linked from the homepage "Who it's for" section.
A full small-business site is now crawled end-to-end instead of stopping at a fixed page limit, and sites without a sitemap are crawled by following links across the whole site.
Re-syncing a bot's website checks what's actually been trained before deciding what to keep, instead of guessing from a text label.
Turn a bot's chat widget on or off from its settings without deleting the bot or losing its knowledge.
The configure screen is now grouped into clear sections โ Identity, Appearance, Behavior & messages, and Access & integrations.
Set how tall the chat panel opens from a bot's settings.
Saving widget settings now shows a reminder to refresh your website so the changes take effect there too.
When a bot can't answer a question, visitors can leave their email right in the chat so you can follow up.
Scale plan. Send captured leads straight into your Salesforce org.
Scale plan. Send lead-capture events to any endpoint you control.
Scale plan. Remove Granetra branding from the chat widget entirely.
Growth & Scale plans. Conversation volume, top questions asked, answer rate, visitor satisfaction, and leads captured, all in one view.
Refresh a bot's website knowledge with one click โ any uploaded PDFs are kept untouched.
Your dashboard now shows conversation usage against your plan's monthly allowance, with an allowed-websites list per bot.
Update your card, change plans, or cancel from a secure billing portal linked in your dashboard.
Every bot now automatically replies in whatever language a visitor writes in, across 95+ languages, with no retraining needed.
Sign-in is now protected against repeated failed login attempts.
Meet the team behind Granetra, and join the new Discord community from the navigation menu.
Homepage FAQ answers now accurately reflect what Granetra can do today.
Append more PDFs or a website URL to an already-trained bot at any time, without starting over.
Customize the first message visitors see when they open the chat.
Upload your own image for the chat widget instead of the default Granetra feather.
Choose a light or dark chat panel, or set it to "Auto" to follow your own website's theme live.
Visitors can rate bot replies with a thumbs up or down, visible per-conversation in your dashboard.
Eased, smoother scrolling throughout the dashboard and the chat widget.
Visitors can attach a text file to the conversation for the bot to answer questions about.
Gradient avatar, a live "Online now" status, and smoother message animations in the chat panel.
Resolved an intermittent "Something went wrong" error during chat.
Embed a trained AI chatbot on your site with one line of code.
Set your bot's tone, accent color, fallback message, and a handoff email for when it can't help.
Fixed click-through and layout issues when the widget is embedded on another website.
Create an account, choose a plan, and train your first bot on PDFs and/or your website in minutes.
Every widget conversation is saved and searchable from your dashboard.