TARA — Killer Apps Catalog

Status: Research draft, 2026-09-16 · Source inputs: founding call transcript (Sriram/Bhavya/Rohan), Vestra.ai use-case taxonomy, Grok Bot product model, OpenMausBot playbook model Constraint set ("the core 5"): web search, web fetch, IMAP email read, scheduled routines (cron), Telegram messaging (+ platform-native persona/memory and the personal microsite).

Strategic frame from the founding call: "The market will figure out which bots it wants if the infrastructure exists." This catalog is not a bet on one bot — it is the seed catalog + demo scripts that let the market reveal itself, and the vocabulary (munshi, sathi, khata) that makes a Docker container feel like a ₹250/month employee.

Critical design constraint discovered while mapping: our toolset reads email (IMAP) but does not send email, and there is no WhatsApp integration we can safely ship (see Traps). Therefore the universal pattern across the whole catalog is "detect → summarize in Hinglish → deliver on Telegram → draft the reply for the owner to send." The bot never acts on the owner's money or customers; it hands the owner a finished draft and a decision. This is both a safety posture and, per Grok Bot's approval-gates, a feature non-technical owners trust.

1. The 18 Bot Ideas

Each entry: promise · routine · integrations beyond the core 5 · perceived value · difficulty (trivial / easy / medium / hard).

Tier A — The email family (founding call: "the killer app is email")

1. Inbox Munshi — "Subah utho, poora email ka saar Telegram pe — kya urgent hai, kya ignore karo."

• Routine: 6:00am + 1pm + 8pm: IMAP sweep of last N hours, classify (payment / customer / government / spam-newsletter), 5-line Hinglish digest on Telegram, urgent items flagged with a drafted reply to copy.
• Integrations: None. Pure IMAP + cron + Telegram.
• Perceived value: Owner stops checking Gmail 15×/day; never misses a GST notice or a payment mail buried under newsletters. Replaces the "office boy who reads email."
• Difficulty: Easy. The flagship. Ship first.

2. Lead Sathi — "IndiaMART/JustDial/website ka enquiry aate hi 2 minute mein Telegram pe, jawaab bhi ready."

• Routine: Every 10–15 min IMAP poll for enquiry emails (IndiaMART "New Enquiry", JustDial leads, contact-form notifications — all arrive by email). Telegram alert: who, what they want, phone number, + a drafted reply in the owner's tone.
• Integrations: None (all lead sources already email). Optional: log to memory for weekly "leads missed" report.
• Perceived value: First responder wins the deal; a trader paying IndiaMART ₹40k/yr for leads loses them by replying next day. This bot alone justifies ₹250.
• Difficulty: Easy (parse known email formats; degrade gracefully to "forward the raw mail").

3. Paisa Kitab — "Roz subah: kal kitna paisa aaya, kiska aaya, kiska atka hai."

• Routine: 7am IMAP sweep of payment-gateway settlement emails (Razorpay/Paytm/PhonePe/Instamojo send structured daily settlements) and bank credit-alert emails → Telegram: "Kal ₹18,450 aaya (7 transactions). Ramesh Traders ka ₹15,000 abhi tak nahi aaya — day 12."
• Integrations: None if scoped to gateway + bank alert emails.
• Perceived value: Reconciliation without opening five apps. The "kitna aaya?" question every owner asks their accountant, answered before chai.
• Difficulty: Medium (bank email formats vary wildly; gateway-only scope makes it easy — recommend launching gateway-only).

4. Vasooli Sathi — "Udhaar yaad rakhta hai, poochhne ka message bhi likh deta hai."

• Routine: Owner tells it in chat: "Ramesh ko 15,000 dena tha 10 tarikh tak." Cron checks daily; on/after due date, Telegrams the owner and offers 3 polite escalating reminder texts to paste into WhatsApp.
• Integrations: None (Telegram + cron + memory).
• Perceived value: Udhaar collection is the #1 silent pain of Indian small business; nobody forgets, nobody feels awkward — the munshi reminds, not the owner.
• Difficulty: Trivial–easy. Massive emotional resonance in sales conversations even if demo wow is moderate.

5. GST Sathi — "GST ki deadline kabhi miss nahi hogi; sarkar ka notice aaya toh turant samjha ke batayega."

• Routine: Standing calendar (GSTR-1 11th, GSTR-3B 20th, QRMP dates) with T-7/T-2/T-0 Telegram nudges; daily web fetch of CBIC press releases/notifications filtered to relevance; IMAP watch for gst.gov.in / incometax notice emails → plain-Hindi "yeh notice kya keh raha hai, kya karna hai, CA ko bhejo."
• Integrations: None.
• Perceived value: Late fees are ₹50–200/day; a CA retainer is ₹1,500–5,000/mo. Positioned as "CA ka peon, CA ka replacement nahi" — files nothing, forgets nothing.
• Difficulty: Trivial. Highest value ÷ effort ratio in the entire catalog.

Tier B — The market-watch family (founding call: "BigBasket price drop")

6. DaamDekho — "Jo saman aap kharidte ho, jab sasta ho — Telegram pe pata chalega."

• Routine: Owner names 5–20 SKUs (BigBasket staples for a kirana-adjacent reseller, supplier listing pages, Amazon/Flipkart product pages). Cron 2×/day web fetch, price-diff against memory, alert only on drops below threshold.
• Integrations: None beyond web fetch; scraping fragility is the risk (BigBasket/Flipkart markup changes, rate-limiting).
• Perceived value: Direct margin. "Bot ne last month ₹3,200 bachaye" is the retention story.
• Difficulty: Medium (per-site fetch adapters; start with sites that behave).

7. Mandi Bhav — "Sona, chandi, gehu, chana, dollar — aapke 8 bhav, roz subah 7 baje."

• Routine: 7am fetch of public rates (data.gov.in Agmarknet mandi prices, gold/silver rate pages, forex, fuel), formatted as a one-screen Telegram card with day-change arrows.
• Integrations: None (public data; Agmarknet has a data.gov.in API).
• Perceived value: For jewellers, grain traders, hardware dealers — replaces calling the mandi or five WhatsApp groups. Niche but daily-habit for that niche.
• Difficulty: Easy.

8. Nazar — "Aapke 5 competitors kya kar rahe hain — har Somvar ka report."

• Routine: Weekly web search + fetch on named competitors: new offers on their sites/listings, rating movement on JustDial/Google, new photos/posts where publicly fetchable. Telegram: "Sharma Electronics ne AC pe ₹2,000 off lagaya; unki rating 4.1 se 4.3 hui."
• Integrations: None.
• Perceived value: Owners are intensely competitive locally and currently learn everything late via word of mouth.
• Difficulty: Easy–medium (search noise needs decent LLM filtering).

9. Review Rakshak — "Koi bura review de, 1 ghante mein pata — jawaab bhi likh ke ready."

• Routine: Every few hours, search/fetch the owner's Google Maps / JustDial listing; detect new reviews; Telegram alert with a drafted reply (humble for bad, grateful for good) in the owner's language.
• Integrations: None, but honesty note: Maps scraping is brittle; official Google Business Profile API needs OAuth (owner's own listing) = phase 2.
• Perceived value: Reputation is the whole business for photographers, clinics, restaurants, coaching. One unanswered 1-star review visibly costs customers.
• Difficulty: Medium (fragility, not logic).

10. Parcel Pal — "Aapke 40 packets mein se sirf wo batao jo atke hain."

• Routine: Owner pastes/forwards tracking numbers once; cron fetches Delhivery/BlueDart/India Post/Ekart public tracking pages 2×/day; Telegram only on exceptions (stuck >3 days, RTO, delivered).
• Integrations: None; courier pages sometimes CAPTCHA — degrade to weekly digest.
• Perceived value: WhatsApp-business sellers shipping daily live in courier anxiety; RTO is pure loss.
• Difficulty: Medium.

Tier C — The growth & habit family (founding call: "inspire them with new ideas")

11. Tyohar Taiyari — "Har tyohar se 10 din pehle: offer ideas, WhatsApp status message, poster text — sab ready."

• Routine: Built-in Indian festival + season calendar (Diwali, Eid, Onam, wedding season, exam season, monsoon for specific trades). T-10/T-3/T-1: trade-specific offer ideas, ready-to-paste Hinglish/regional customer messages, stock-up reminders ("pichhle saal is hafte demand badhi thi").
• Integrations: None. LLM + calendar + memory of the owner's trade.
• Perceived value: Owners know they should do festival marketing and never get to it. Best demo in the catalog: ask it live, get a Diwali offer for their shop in 10 seconds.
• Difficulty: Trivial. Bundle into every pack.

12. Udaan — "Roz subah ek naya idea, aapke business ke liye, aapki bhasha mein."

• Routine: 7:30am: one concrete, small, actionable idea filtered to the owner's trade + city + what it remembers of past conversations; weekly "idea of the week" with a worked example from a similar business.
• Integrations: Web search optional (ground ideas in current trends).
• Perceived value: Straight from the founding call — non-technical owners want an assistant who inspires them. This is the retention/love bot, not the acquisition bot.
• Difficulty: Trivial. Near-zero marginal cost. Ship inside every other bot as a morning sign-off line.

13. Trade Chai — "5 minute mein: aaj ki 4 khabrein jo aapke dhanda affect karti hain."

• Routine: 7am web search filtered by trade + city ("electronics retail", "Ludhiana hosiery"), summarized ruthlessly; skip politics unless regulatory.
• Integrations: None.
• Perceived value: Makes the owner the best-informed person at the mandi/market. Habit-forming.
• Difficulty: Trivial.

14. KaamYaad — "Bol do, bhool jao — sahi waqt pe yaad dila dunga."

• Routine: Natural-language reminders in chat ("har Friday supplier ko call", "Tuesday ko Ramesh ka payment check"), recurring checklists, and — meta-feature — the owner creates their own routines in plain words ("roz raat 9 baje mujhe din ka summary bhejo"), which the platform stores as new cron entries. This is how "the market figures out which bots it wants" — watch what routines owners create.
• Integrations: None.
• Perceived value: The daily-use glue; the bot that feels most like an employee.
• Difficulty: Trivial — and it's the telemetry goldmine for the whole product.

15. Kagaz Samjho — "English ka notice/letter/invoice photo kheencho, Hindi mein samajh lo."

• Routine: On demand: owner forwards a photo/PDF on Telegram → bot reads it (multimodal LLM) → "Yeh kya hai, kya maang raha hai, deadline kab hai, kya karna chahiye." Also works for supplier price-lists and English contracts.
• Integrations: Needs Telegram file receive (already implied by chat surface) + a cheap vision model.
• Perceived value: Enormous for owners with weak English: government notices, bank letters, legal threats all become readable. Demo wow is the highest in the catalog.
• Difficulty: Medium (vision-model cost/quality tuning; hallucination guardrails for legal-ish text — always append "confirm with your CA").

Tier D — The books & web presence family

16. Khata Munshi — "Kharcha bol do, Sunday ko poora hisaab mil jayega — CA ko forward kar do."

• Routine: Owner voice-notes/texts expenses and sales in chat anytime ("aaj 2400 ka maal liya", "scooter ka petrol 300"); Sunday 7pm weekly P&L card; month-end export formatted to forward to the CA. This is "managed accounting lite" — the founding call's karta/bala/munshi idea, deliberately kept short of filing (see Traps).
• Integrations: None (memory + cron). CSV/PDF export = nice-to-have.
• Perceived value: Owners currently keep hisaab in a diary or their head; the month-end CA scramble disappears.
• Difficulty: Medium (messy natural-language entry needs careful confirmation loops — "aapne ₹2,400 likha, sahi hai?").

17. Dukaan.site — "Aapki dukaan ka chhota website jo customer se khud baat karti hai."

• Routine: Standing, not scheduled: the product's built-in microsite becomes a talking shop-front — catalog, hours, directions, FAQ answered by the bot in character; weekly Telegram report: "Is hafte 47 log aaye, 6 ne price poocha, 2 ne call kiya."
• Integrations: None (microsite + analytics are platform-native).
• Perceived value: "Meri dukaan ka website hai aur wo customers se baat karti hai" — status + utility. Strong wow in demos to owners' peer groups.
• Difficulty: Medium (mostly already built as product surface; incremental is the chat persona + weekly report).

18. Sarkar Scout — "MSME subsidy, naya scheme, relevant tender — aapke liye ho toh bata dunga."

• Routine: Weekly web search/fetch of PIB, MSME ministry, state industry portals' news/press-release pages; match against owner profile (turnover band, sector, state); Telegram only on strong matches with "kaise apply karein — apne CA se poochho" pointer.
• Integrations: None. Strictly alert-only — no portal logins, no form filling (see Traps).
• Perceived value: Feels like insider access; owners suspect they're leaving government money on the table (they usually are).
• Difficulty: Easy (alert-only); becomes a trap the moment it touches application workflows.

2. Ranking — (demo wow × willingness to pay) ÷ build difficulty

Scores 1–5; difficulty: trivial=1, easy=2, medium=3, hard=4. Score = (wow × pay) / diff.

Ranking caveat: scores assume a generic owner. Per segment the order reshuffles — Mandi Bhav is #1 for a jeweller, Parcel Pal top-3 for a WhatsApp seller, Kagaz Samjho top-3 for a CA-firm peon-turned-operator. Sell segment packs, not a menu (see §5).

3. Top 5 — need NOTHING beyond web search, web fetch, IMAP, cron, Telegram

All five are buildable on the current stack with zero new integrations, and all five follow the safe "detect → summarize → draft" pattern:

1. Inbox Munshi — IMAP + cron + Telegram. The wedge. Every activation demo starts here.
2. Lead Sathi — IMAP + cron + Telegram. The money bot; makes ₹250 feel like theft (in our favor).
3. GST Sathi — cron + web fetch (CBIC) + IMAP (notice watch) + Telegram. Universal pain, trivial build.
4. Tyohar Taiyari — cron + LLM + Telegram. The demo-party trick and the viral-share bot.
5. Vasooli Sathi — Telegram + cron + memory. The emotional closer: everyone is owed money.

Runner-up (honorable mention): KaamYaad — same zero-integration profile; include it free inside every bot because owner-created routines are the discovery engine for the next catalog.

A v1 bundle of just these five = "aapka ₹250/month office boy: email padhta hai, leads batata hai, GST yaad dilata hai, tyohar pe marketing likhta hai, udhaar nahi bhoolta." That sentence alone can carry the first 100 customers.

4. Trap Ideas — sound great, will hurt us

1. WhatsApp auto-read / auto-reply ("WhatsApp Munshi"). The most-requested, most dangerous. Unofficial APIs = number bans (the owner loses their business number — existential); official WhatsApp Business Cloud API = Meta verification, per-conversation fees, template approvals, 24-hour windows, and since Jan 2026 a ban on open-ended AI assistants. Safe alternative shipped today: every bot drafts WhatsApp messages as copy-paste text (Vasooli, Tyohar, Lead Sathi all do this). Revisit only via an official BSP partnership, post-PMF.
2. "AI accounting that files your GST" (full Munshi). Filing needs GSP integration with GSTN, authentication/EVC, and carries liability for wrong returns — one wrong filing and word-of-mouth kills us in a tightly-networked trader community. Safe version: Khata Munshi + GST Sathi = reminders, records, drafts; the human CA signs. Sell as "CA ka peon," never "CA ka replacement" (also keeps CAs as a referral channel instead of enemies).
3. Auto-purchasing / auto-payments ("bot buys when price drops," "bot pays suppliers"). Payment initiation puts us in RBI PA-PG/PPI territory, needs mandates/tokenization, and one fraud incident is fatal at this price point. Safe version: DaamDekho alerts; the human buys.
4. Government-portal automation (GST portal login, e-way bill, GeM bidding, IT portal). OTP/CAPTCHA/session automation against portal ToS, extreme brittleness, and regulatory blast radius. Safe version: Sarkar Scout and GST Sathi watch public pages and email notices only. Alert, never act.
5. Bank/net-banking scraping for "real-time balance & UPI monitoring." Storing banking credentials is a compliance and trust cliff; the legitimate path (RBI Account Aggregator) is a partnership marathon. Safe version: Paisa Kitab reads the bank's own email/SMS-bridge alerts via IMAP — same outcome, zero credential risk.

Honorable trap mention: bulk WhatsApp marketing blasts (spam bans + TRAI telemarketing rules) and Instagram DM automation (ToS, account loss). Both will be requested by customers; both get the draft-don't-send answer.

5. What the reference products teach (filtered for our audience)

Vestra.ai (Gurugram, ~$20/mo, "agents for every team"): take package by role, sell an outcome ("come back to done") — our segment packs are the Indian-SMB mirror of their per-team pages. Filter: their buyers have CRMs and Slack; our buyer's "CRM" is a diary and their "Slack" is a WhatsApp group — everything must degrade to email + Telegram + copy-paste drafts.

Grok Bot (xAI; ~$30/mo): take "narrow specialised bots > one universal assistant," "finished work, not advice," "bots stop and ask before acting." Routines are the atomic unit — sell routines, not "agents." Filter: ₹250/mo only works with cheap open models, which is fine because none of the top 5 need a frontier model.

OpenMausBot (open-source Grok Bot): take the playbook/pack install format — persona + routine packs installed with connections OFF and routines PAUSED, each flipped on with one tap. That is our marketplace blueprint and it directly answers "the market will figure out which bots it wants."

India-context filters applied throughout: UPI behavior → read settlements, never initiate payments. WhatsApp-first → draft for WhatsApp, never touch the API. GST → calendar + notice-watching, never filing. Festival/wedding seasonality → first-class bot, not a feature. Language → Hinglish/regional register by default.

6. Packaging — sell packs, not a menu

First 100-customer motion: default-install Inbox Munshi + GST Sathi + Tyohar Taiyari + KaamYaad for everyone (all zero-extra-integration), demo Kagaz Samjho live for the jaw-drop, and let KaamYaad's owner-created routines tell us what to build next. That's the market figuring out which bots it wants — instrumented.

POC personas (locked, D008)

The POC ships three celestial personas + a baseline companion (naming per D034): VEGA (Virtual Email Guardian Assistant — inbox digest, was Inbox Munshi), NOVA (Notifications Of Valuable Advice — daily idea, was Udaan), PULSAR (Price Updates & Low-price Signals, Alert Radar — price watch, was DaamDekho), and SIRIUS (baseline companion, was Saathi) — covering the three families with zero extra integrations. Catalog entries keep their Hindi archetype names until they are promoted to shipped personas.