| Agent builder |
Live |
Agents |
| How smart each agent is: Smart, Smarter or Smartest |
Live |
Smart (the default) costs least; the builder shows about what a reply costs on each. All three are Anthropic Claude models |
| Backup model |
Live |
Retries a reply once on another model when the main one is down. Advanced settings |
| Variables in instructions and the welcome message |
Live |
Up to 30 per agent. Variables |
| Reply length and reply language |
Live |
Short, medium or long; English, Hindi, Hinglish or the customer's language |
| Suggested prompts and a usage guide in the test chat |
Live |
|
| Long-term memory |
Live |
Notes kept between conversations; new notes from the agent ask first. Long-term memory |
| Task fields |
Live |
Details recorded on each task, shown on the task and in exports. Task fields |
| Hide contact details from Viewers |
Live |
In Tasks, exports, leads and memory |
| Knowledge pasted as text, with a search tool |
Live |
Knowledge |
| Teach your agent by chatting |
Live |
No website needed: answer its questions, or send a photo of your rate card, a PDF or a link, in Hindi, Hinglish or English. Each fact waits on a "Here's what I learned" card until a Builder or above accepts it; customers can never teach. Teach your agent by chatting |
| Your business, written once for every agent |
Live |
Details, how agents should talk, and what they must never say or promise. Your business |
| Search knowledge by meaning |
Live |
Finds the right part even when the words differ, in English, Hindi and Hinglish, blended with word matches. Search by meaning |
| Hide ID and card numbers from the AI |
Live |
On by default. Aadhaar, PAN, card and bank numbers and OTPs are replaced before the AI reads a message. What the AI never sees |
| Built-in tools |
Live |
Tools |
| Approval setting for each tool |
Live |
Runs freely, Ask first, or ask and then run after a waiting time |
| Autonomy limit |
Live |
1 to 50 steps per message |
| Test chat |
Live |
Nothing is really sent. Each reply shows its steps and cost |
| Evaluations: test sets, checks and scores |
Live |
Runs in test mode, with CSV import, AI-written cases and run comparison. Evaluations |
| Must pass before publishing |
Live |
Admins can publish anyway with a reason, kept in the audit log. Evaluations |
| Find the cheapest level that passes |
Live |
A test set on Smart, Smarter and Smartest at once, then switch in one click. Evaluations |
| Daily quality checks on real conversations |
Live |
Off until you turn it on; each checked conversation costs a little. Live quality |
| Publish, version history, pause and resume |
Live |
|
| Make an earlier version live again |
Live |
Agents |
| Follow up when a customer goes quiet |
Live |
At most two emails, checked first unless you choose otherwise. Follow up quiet customers |
| Read a web page and search the web |
Live |
Up to 8 pages and 5 searches a task. Tools |
| Agents list: search, filters, sort, and list or grid view |
Live |
Find and organise agents |
| Folders for agents |
Live |
Deleting a folder keeps its agents |
| Duplicate an agent, or copy it to another project |
Live |
Copies are drafts. Knowledge files stay in their project |
| Export and import agents |
Live |
A JSON file of the settings. Knowledge files and app actions are listed, not included. Export and import |
| Replies in English, Hindi or Hinglish, with AI disclosure |
Live |
|
| Leads page |
Live |
One per project, under Operate: a board by stage (New, Qualified, Call booked, Won, Lost, Not a fit) and a list, a Today strip, filters, search and CSV export. One lead per person: the same email or phone attaches to the same lead. Test leads are kept apart and marked. Leads |
| Lead qualification |
Live |
Up to 5 criteria per agent, from its template or your own. Each is answered with the customer's own words, kept only if they appear word for word; the verdict and 1 to 5 score are worked out from the weights, not by the AI. How a lead is checked |
| Cost per lead |
Live |
Each lead shows the credits its conversations used. Analytics → Leads and cost gives cost per lead, qualified lead, booked call and won client, by source, agent and week. Cost per lead |
| Merge duplicate leads |
Live |
Owners and Admins merge two or more leads from the list or a lead. A preview picks the newest value per field and you can change any; the oldest lead is kept and the others point to it, so nothing is lost. Merge duplicate leads |
| Lost reasons and suggestions to close |
Live |
Lost asks why (not a fit, no budget, chose someone else, no reply, other), counted in Analytics. The AI suggests closing leads that aren't a fit or didn't reply after 2 follow-ups; your team closes them, one by one or in bulk. A Lost lead that writes again is flagged Came back. Suggestions to close |
| Morning brief |
Live |
9:00 each morning to Owners and Admins, per project: what your AI team noticed, then new and qualified leads, calls today, follow-ups and leads waiting over a day. No contact details or quotes; it never writes anything and isn't charged. Each person can turn it off on the Leads page. The morning brief |
| What your AI team noticed, on Home |
Live |
Up to three things worth doing today, such as a lead waiting a day for a reply, a qualified lead with no call booked, or a conversation handed to your team. Found by rules on your own data, with no AI. A button that writes shows its price first and waits for your OK; Not now hides a card for a week. What your AI team noticed |
| Meeting notes feed leads |
Live |
A notetaker meeting with a lead (from its booking's Meet link, or its email in the meeting) is added to the lead and checked; suggested changes, with word-for-word quotes, wait for you to accept or reject each. Meetings with a lead |
| Time to first reply |
Live |
On each lead, and the average by where leads came from in Analytics → Leads and cost |
| IndiaMART leads |
Pilot |
Through a webhook with IndiaMART's Push API: tested with IndiaMART's documented fields, not yet with a live IndiaMART account. Get IndiaMART leads |
| Next step drafted and sent for each lead |
Live |
The AI drafts each lead's next email; you approve, edit or skip it on the Leads page or in Tasks. It goes from your Gmail, or the agent's own address, only to that lead and in sending hours: at most 2 follow-ups per lead, and a reply or STOP stops them. Next step |
| Speed-to-lead |
Live |
A first reply to a new form, HubSpot, webhook or API lead within minutes, offering free times when the agent books calls. It waits for your OK, or sends itself at any hour if you choose. Speed-to-lead |
| Reminders before a booked call, and no-shows |
Live |
An email to the lead before the call, with the Meet link. A call deleted or declined on the calendar moves the lead back to Qualified; after a no-show, the AI drafts a message offering another time. Reminders |
| Next steps by WhatsApp |
Planned |
Once your workspace can send from its own WhatsApp number |
| Zoho CRM and Bigin |
Planned |
For now, export a CSV or use the "A lead changes stage" event with a web request |
| Knowledge files: PDF, Word, Excel, CSV, TXT and Markdown uploads, and notes |
Live |
Up to 4 MB a file; we keep the text, not the file. Knowledge |
| Calendar |
Live |
A week of scheduled runs, calls booked and posts published, in India time, with every repeating schedule. See your AI team's week |
| Goals |
Live |
Pick what you want help with on Home, and your kind of business if you like, which picks the template that fits it; we set up the agents, tools and schedule it needs, and a checklist shows what's live, what needs you, what's waiting on another step and what isn't available yet. Quickstart |
| Post to LinkedIn |
Live |
Your own profile, after you approve each post, with a stock photo or your own if you like. Post it now or schedule it for a time you pick. Company pages need LinkedIn's approval, which we haven't applied for yet. Reconnect every 60 days. Post to LinkedIn |
| Post to Instagram |
Pilot |
Photo posts to a professional (Business or Creator) Instagram account, after you approve each one. Post it now or schedule it for a time you pick. For now only accounts our team adds to our Meta app; every account can connect once Meta has reviewed our app, which we haven't applied for yet. Until then, each Instagram post comes as a photo and caption with Share and Save. Post to Instagram |
| Stock photos for posts |
Live |
Free photos from Pexels, with the photographer's credit. Images for your posts |
| Set up an agent from your website |
Live |
Paste your address: we read up to 25 pages, pick the template, and make a draft agent with tests to try before you publish. Quickstart |
| Template gallery |
Live |
16 templates and a blank agent, including discovery call booking, an agency enquiry desk, an appointment setter for warm leads, real-estate site-visit booking, D2C product questions, answering from your website, and festival and weekly posts for a shop. Each has at least 6 ready tests, qualification questions where it takes leads, example conversations in English and Hinglish, and an estimate of a month's cost. Filter by kind of business in the app, and by goal too on the website. Each comes with an invented example business, loaded only if you tick Start with example data. In the app, open Agents, then Templates. Templates |
| Team chat: talk to your live agents in the app |
Live |
Everyone in the workspace, viewers included. Chat with your agents |
| Photos and PDFs in the test chat, Team chat and the API |
Live |
Up to 3 a message, 4 MB each, PDFs up to 20 pages, kept for 90 days. The agent's Let people send files setting. Send photos and PDFs |
| Run an agent in bulk from a CSV |
Live |
Up to 500 rows, as a test or a real run, with a results file. Run an agent in bulk |
| Website import for knowledge |
Live |
One page or a whole site, up to 50 pages an import and 500 a project, with weekly or monthly refresh. Knowledge |
| Knowledge tables |
Live |
Up to 5,000 rows each, typed in or imported from CSV or Excel. Agents look up rows with the Look up a table tool. Tables |
| AI columns in tables |
Live |
Filled by an AI prompt or one of your agents, with the rows and cost shown first. Fill a column with AI |
| Describe a task and get a draft agent, and Write with AI for instructions |
Live |
Drafts are never published by themselves |
| Your own tools: web request, AI, app action, condition and format steps |
Live |
Build without code, test with each step's details, publish with versions. Tools |
| Your own tools in agents, as a form or through the API |
Live |
Tools that change data ask first. Running your own tools |
| Tool secrets |
Live |
Encrypted, never shown again, and can be limited to certain sites. Tool secrets |
| Run your own tools over many rows at once |
Planned |
|
| Workforces: teams of agents that hand conversations to each other |
Live |
Canvas, conditions, hand-overs with a safety stop, test mode, publishing with pinned agent versions, and version history. On a phone, a read-only list. Workforces |
| Workforces in Team chat, the run API, schedules and webhooks |
Live |
Run a live workforce |
| Website chat and share links for a workforce |
Live |
Set up in the workforce's Run tab. Share a chat link |
| Test sets for a whole workforce |
Live |
With checks for which agent answered and whether a hand-over happened. Test sets for a team |
| Marketplace |
Planned |
|