# BitFoundry and WorkWeaver: the full record

BitFoundry Tech Private Limited (brand: BitFoundry) is a software product company. It was founded in 2025 and is based in Coimbatore, Tamil Nadu, India (CIN: U62011TZ2025PTC035159). It is a member of the NVIDIA Inception Program. Its only product is WorkWeaver. Company contact: contact@bitfoundry.ai. Product contact: workweaver@bitfoundry.ai. Founders: Aditya Koushik V (Co-founder and CEO, https://www.linkedin.com/in/aditya-koushik-v-33b35482/) and Santhanakrishnan Dhandapani (Co-founder and CTO, https://www.linkedin.com/in/santhanakrishnan-dhandapani-156418184/, https://x.com/Santhanakrishna). WorkWeaver is hosted on Amazon Web Services today and is not locked to any one cloud.

BitFoundry is a software product company. We build WorkWeaver. WorkWeaver is one product, the same for every customer, priced by use: USD 9 per agent-hour, priced by the minute (USD 0.15 per agent-minute), with no seats, no minimum, and the first 7 days free from the first run. Models, calls and WhatsApp run on the customer's own keys and accounts, with no mark-up.

## What WorkWeaver is

WorkWeaver gives a company agents that learn how it works and run it around the clock while its people lead. Agents carry every handoff, so work stops waiting between people and teams. People were never meant for the labour of chasing, copying and waiting; they are meant for the design: judgment, taste, direction and vision. WorkWeaver plans the job, does the work in the tools a company already owns, shows why it acted, and keeps what the company learned.

Human speed is 40 hours of work a week. Agentic speed is 168 hours a week. The old way of working: know-how leaves with people and new hires take six months to catch up; everyone hits their own target while the company number does not move; work waits between people for days in inboxes and nobody owns the wait; growth means more coordinators and headcount grows faster than output; the best people chase and copy and their day goes on work about work. The new way of working: know-how stays with the company and new hires work its way from day one; every goal links to the company goal and everyone pulls the same way; agents carry every handoff, any hour, the same day; growth without the overhead; the best people are with customers, building relationships and trust.

## Every capability

Company memory. One memory the whole company writes to, filled from the meetings, messages, documents and decisions it already has. Every team writes to it, and every agent uses it. When two things disagree, the newer decision wins and the old one is kept with its date. Personal, team and private vaults exist, and each person sees only what they are allowed to see. Every answer shows where it came from. Export every memory any time.

Company-wide skills. A skill is how a company does one job: the steps, the rules and the judgment calls. What one person teaches becomes a skill the whole company uses, from the first day. Skills are written from real work and real edits, so nobody has to write a manual. What one team teaches, every team uses. New hires work the company's way from day one, and know-how stays when people move on. Every version is kept with who taught it and why. Roll back in one click.

Better every week (the quality loop). Every change is tested on a company's real past work first, so small mistakes are caught before they pile up. Every skill has a health score made of checks people can read. Independent judges grade the work, and no model grades its own work. A change goes live only if it helps on past work and the company goal. Every version is kept.

No new app (every channel). WorkWeaver joins meetings on Zoom, Google Meet and Microsoft Teams, takes notes and answers out loud. It calls in and out from its own phone number, on the company's phone account, and sends from its own email address. It answers on the company's WhatsApp Business number. Approvals arrive in Slack and Teams. It connects to every tool a team runs on: CRM, accounting, documents, calendar and project tools.

Any model at the lowest cost. WorkWeaver tests cheaper models on a company's own past work. If one scores as well, it recommends the switch and the company applies it in one tap. A model that scores lower never goes live. Each task uses the model that scores best on it, and if one fails another takes over on its own. Models run on the company's own keys and the provider bills the company, with no mark-up from us.

Works inside the tools each team already opens. The common core across every team is email, chat, meetings, sheets and docs: Microsoft 365, Google Workspace, Gmail, Outlook, Slack, Microsoft Teams, Zoom, Google Meet, WhatsApp, Excel, Google Sheets, Notion, Jira and Confluence. Each team also has its own tools. Marketing: HubSpot, Google Analytics, Google Ads, Meta Ads Manager, LinkedIn Campaign Manager, Semrush, Ahrefs, Figma, WordPress, Webflow, Mailchimp, Hootsuite, MoEngage and Zoho Campaigns. Sales: Salesforce, HubSpot Sales Hub, Zoho CRM, Freshsales, LeadSquared, Pipedrive, LinkedIn Sales Navigator, Apollo.io, ZoomInfo, Gong, Outreach, PandaDoc, IndiaMART and Exotel. Onboarding: Rocketlane, GUIDEcx, Arrows, Smartsheet, Loom, Intercom, Pendo, WalkMe, Userpilot, Gainsight, Digio, Signzy and HyperVerge. Customer relationships: Zendesk, Freshdesk, Zoho Desk, Intercom, Salesforce Service Cloud, HubSpot Service Hub, Gainsight, ChurnZero, Planhat, Gorgias, Gupshup, WATI, Ozonetel and Twilio. Product development: GitHub, GitLab, Linear, Figma, Postman, Sentry, Datadog, AWS, Google Cloud, Vercel, Amplitude, Mixpanel, PagerDuty and GitHub Copilot. IT: Microsoft Entra ID, Okta, JumpCloud, Microsoft Intune, Jamf, Jira Service Management, Freshservice, ManageEngine, ServiceNow, 1Password, Bitwarden, CrowdStrike and Vanta. Hiring: LinkedIn Recruiter, Naukri, Indeed, Greenhouse, Lever, Ashby, Zoho Recruit, Keka, Darwinbox, greytHR, BambooHR, Deel, HackerRank and AuthBridge. Legal: Microsoft Word, Adobe Acrobat Sign, Ironclad, SpotDraft, Juro, LinkSquares, Leegality, Digio, Zoho Sign, SCC Online, Manupatra, Westlaw, LexisNexis and Clio. Operations: Smartsheet, ClickUp, Airtable, Make, n8n, Lucidchart, Miro, Zoho Creator, Odoo, Tally, Shopify, Shiprocket, Unicommerce and NetSuite. Finance: NetSuite, Tally, Zoho Books, BILL, Tipalti, Ramp, Expensify, SAP Concur, Coupa, Stripe, Razorpay, Cashfree, Chargebee and HighRadius. Accounting: QuickBooks, Xero, TallyPrime, Zoho Books, Sage Intacct, Busy, Marg, FloQast, BlackLine, Numeric, Workiva, ClearTax, Avalon and Dext. A missing tool is connected during the first two weeks of a guided rollout.

Humans on the loop. One agent carries the work across every department, from step to step. People teach (show it once), trust (give more room step by step), judge (make the calls), set the taste (define what good is) and monitor (step in when it matters). Trust grows step by step, one kind of task at a time: first it drafts and a person sends; then it does the work and asks before each risky step; then it does the work and reports after; then it runs the work inside the limits set for it. When someone edits a draft, it learns why. One agent sees the whole company, so it finds the one bottleneck that moves the company goal, at agentic speed.

## Security and control

WorkWeaver asks before risky steps. The company sets where it must ask: by amount, customer, supplier, kind of action and team. It never moves past that line on its own. Approvals arrive where people already are: Slack, Teams, email or WhatsApp, in one tap. Trust grows step by step: limits widen when the record earns it. One button stops every agent at once.

Example approval rules, set by the company: pay a known supplier under USD 5,000 and the agent does it and reports after; pay any amount to a new supplier and the finance lead approves first; refund over USD 500 and the support lead approves first; change bank details and only people can do it, never agents.

Every action is on record: what it did, when, why, and who said yes. Nothing happens off the books. Each action is logged with its reason and the facts it used, and every approval is logged with the person who gave it. Search the trail by customer, person, tool or day, and export it any time. A sample Thursday trail: 02:14, matched 46 supplier bills to orders under the three-way match rule taught by Anil; 06:30, paid 12 known suppliers, USD 31,200 in all, each under USD 5,000, reported to finance; 07:45, asked to pay a new supplier USD 18,400 under the new-supplier rule; 08:02, approved by Anil, payment sent, approval kept with the payment.

WorkWeaver has its own name and its own access. It never acts as one of your people. It has its own email address and its own phone number on your phone account. Access is set per tool and per action: read, write, or none. CRM read and write, accounting software read and write, bank pays approved bills only, payroll no access. Each person sees only what they are allowed to see in memory and in reports. Remove access in one place and it is gone everywhere.

Company data stays with the company. The company's workspace is yours alone. Export every memory, skill and audit record in open formats, any time. Shared models are never trained on company data, and model providers used with the company's own keys follow the company's agreement with them. Leave any time: take everything, and the rest is deleted.

Accounts stay with the company. Phone numbers and calls, WhatsApp messages and AI models run on the company's own accounts and keys, and each provider bills the company directly, with no mark-up. The product is cloud agnostic and not tied to any cloud provider.

Written commitments are published at https://workweaver.ai/privacy (Privacy Notice), https://workweaver.ai/terms (Terms of Service) and https://workweaver.ai/dpa (Data Processing Addendum).

## Pricing

USD 9 per agent-hour, priced by the minute. You pay only for the minutes agents work. No rounding up, no minimum, no seats. The meter runs while the agent works. USD 0.15 per agent-minute, added up over the month. A 90-second task costs USD 0.225. Four working minutes on one supplier bill make four agent-minutes.

Three worked examples, counting working minutes only. A finance team checking 150 supplier bills a month: 150 bills times 4 agent-minutes of work per bill equals 600 agent-minutes, which is 10 agent-hours, so USD 90 a month. A clinic confirming 18 patients a day: 18 unconfirmed patients times 1.5 agent-minutes times 22 working days equals 594 agent-minutes, which is 9.9 agent-hours, so USD 89.10 a month, plus phone minutes billed by the clinic's own phone provider. A sales team answering 300 new enquiries a month: 300 enquiries times 3 agent-minutes equals 900 agent-minutes, which is 15 agent-hours, so USD 135 a month.

WorkWeaver bills agent-hours at USD 9 each, with no seats and no minimum. The customer's own providers bill phone numbers and calls (Twilio), WhatsApp messages (WhatsApp) and AI models (the customer's model provider). No mark-up from us.

The first 7 days are free, from the first run. Models and WhatsApp run on the customer's own keys, so the customer pays WorkWeaver nothing that week. Then the meter starts. After your 7 free days, the meter starts at USD 9 an agent-hour.

The guided 90-day rollout is setup and training for WorkWeaver; agent work is billed at USD 9 an agent-hour.

Pricing questions, answered. What counts as an agent-hour? Only the minutes agents work: USD 0.15 an agent-minute, added up over the month; four working minutes make four agent-minutes and sixty make an agent-hour. When does the meter run? While the agent works; a 90-second task costs USD 0.225. Is there a minimum? No: no minimum and no seats, so a quiet month costs less. What do models and phone calls cost? Whatever the customer's providers charge, with no mark-up.

## All 16 workflows (examples from the WorkWeaver use-case library)

The figures below are examples from the WorkWeaver use-case library, not results of any named customer.

1. Sales and production: custom quotes wait nine days. Before: a law firm asks on WhatsApp for 60 desks with custom trays; the quote waits 4 days for engineering, an engineer prices it from memory at night, then it waits 5 days for finance; the quote goes out on day 9 and 4 in 10 buyers have already bought elsewhere. Under an hour of real work spread over 9 days of waiting. After: the agent drafts each custom quote from 1,200 past quotes and checks it against the finance rules; approvers check the margin in one tap. Quotes in the pile: 34 to 0. Evenings spent pricing: most nights to none. A custom quote: 9 days to under a day. https://workweaver.ai/workflows/custom-quotes/
2. Finance: supplier bills take 18 days to pay. Before: the bill lands in a shared inbox with 60 others, waits, is typed in, waits on the warehouse and the buyer, the price is wrong and boxes are crushed, the owner approves in 2 minutes, then it waits for Thursday's payment run; paid on day 18. 44 minutes of real work spread over 18 days. After: the agent checks every bill against the order and the delivery and asks the supplier to fix what is wrong, with proof. Bill to paid: 18 days to 2 days. "Where is our money" calls: 2 to 0. Early-payment discount: missed to USD 96 kept. https://workweaver.ai/workflows/supplier-bills/
3. Operations: one in five appointments is a no-show. Before: a reminder text goes out, 18 of 90 do not reply, the front desk calls between check-ins and reaches 9 of 18, and 17 patients do not come; 3 hours of front-desk calls a day and the waitlist is never called. After: the agent calls every patient who has not confirmed, moves those who cannot come and offers the freed chairs to the waitlist. Empty chairs on a Wednesday: 17 to 4. Front-desk time on reminder calls: 3 hours a day to none. Waitlist patients seen: 0 to 6. https://workweaver.ai/workflows/no-shows/
4. Marketing: a campaign takes three weeks to launch. Before: the brief goes by email, drafts wait 4 days, legal has not seen anything, legal strikes claims, two more rounds by email; live on day 21, two days after the sale weekend. About 5 hours of real work spread over 21 days. After: the agent turns the brief into a plan with an owner and a date for each piece and checks every draft against the brand rules and past legal decisions. Brief to launch: 21 days to 4 days. Rounds of changes: 6 to 2. The sale weekend: missed by two days to live two days early. https://workweaver.ai/workflows/campaign-launch/
5. Sales: new leads wait three days for a reply. Before: a 40-clinic group asks for a price on Friday night and waits all weekend; a rep calls Monday, voicemail, seven emails to book a call; on the call the price needs a manager who is at an offsite; price sent day 15 and the buyer went with another tool. 50 minutes of real work spread over 15 days. After: the agent answers every enquiry within minutes, nights and weekends included, with the group price and two meeting times. First reply: 65 hours to 2 minutes. Price in the buyer's hands: day 15 to day 1. The deal: lost to a faster vendor to proposal on day 5. https://workweaver.ai/workflows/new-leads/
6. Onboarding: new customers wait six weeks to go live. Before: kickoff call, 14 decisions go into one person's notes, nobody owns the next step, three reminders to the customer's IT, the file arrives in the wrong format, training on day 44. About 6 hours of real work spread over 44 days. After: the agent joins the kickoff, posts every task with an owner and checks each file the hour it arrives. Signed to live: 44 days to 12 days. Reminders the onboarding lead sends: 11 to 0. https://workweaver.ai/workflows/customer-onboarding/
7. Customer relationships: customers explain their problem three times. Before: a cracked bowl, auto-reply of 2 business days, asked for photos, she tells the story again, waits on stock checks, calls on day 6, a new agent cannot find the thread, refund approved at 7 pm on day 9 with no mixer. 25 minutes of real work spread over 9 days and three tellings. After: the agent answers in the customer's language, asks for one photo and books the fix while she is still on the chat. First real answer: day 2 to 2 minutes. Times she told the story: 3 to 1. Solved day 9 with refund only to day 3 with a new mixer. https://workweaver.ai/workflows/customer-complaints/
8. Customer relationships: renewals are noticed a week before they lapse. Before: the champion leaves, the new safety head never logs in, logins halve, two support tickets wait a week, the account still shows green, the renewal reminder goes out 7 days before and the customer has chosen another provider; USD 84,000 a year gone. The warning signs were there four months early and nobody was looking. After: the agent reads the signs spread over five tools and tells the account manager months ahead. Risk noticed: 7 days before renewal to 120 days before. The renewal: lost to renewed for 2 years plus 2 sites. https://workweaver.ai/workflows/renewals/
9. Product development: customer bugs wait nine days for an engineer. Before: a studio reports members cannot book classes on their phones, support asks for steps and a screenshot, cannot repeat it, waits for Tuesday's triage, triage asks which phones, an engineer picks it up on day 9: a two-line fix. 40 minutes of real work spread over 9 days. After: the agent gets the details the first time, repeats the bug and finds every studio it hits. Report to engineer: 9 days to 1 hour. Times the studio was asked for details: 3 to 1. Studios told it was fixed: 1 of 11 to all 11. https://workweaver.ai/workflows/bug-handoff/
10. IT: access to one app takes four days. Before: a new account manager asks in Slack, is told to raise a ticket, IT needs her manager's approval, the manager is travelling, 46 tickets sit in the queue after the weekend, no spare licence, waits on finance; access granted Monday 4:40 pm and a customer quote went out three days late. 12 minutes of real work spread over 4 days. After: the agent knows each person's role and IT's rule for every app; the manager taps yes and access is set up with the approval logged. Request to access: 4 days to 20 minutes. Days a new account manager could not quote: 3 to 0. https://workweaver.ai/workflows/access-requests/
11. Hiring: hiring takes 57 days and day one has no laptop. Before: the job goes live, 180 CVs wait two weeks in a pile, first interview after six emails, waits for the boss's calendar, she says yes on day 45, nobody tells IT, first day with no laptop and no logins. 3 hours of real work spread over 57 days. After: the agent reads every application against the hiring manager's list, books the calls, drafts the offer and sets up payroll, the laptop and accounts. Opening to a yes: 45 days to 5 days. First day: no laptop and no logins to logged in at 9:05. https://workweaver.ai/workflows/new-hires/
12. Hiring: four interviewers, three weeks to find one slot. Before: the candidate passes round one, the final needs 4 interviewers, the recruiter emails all four, a slot is found then an interviewer travels, rebooked, 31 emails, feedback from 2 of 4, waits 4 days for the rest, offer ready day 23 but she took another job on day 20. 4 hours of real work spread over 23 days. After: the agent finds one slot in four calendars, briefs each interviewer and chases every feedback form within the hour. Final round held: day 19 to day 4. Scheduling emails the recruiter wrote: 31 to 0. The candidate: took another offer on day 20 to said yes on day 6. https://workweaver.ai/workflows/interview-loop/
13. Legal: contracts take three weeks to sign. Before: the customer says yes, sales edits an old contract, waits on their lawyer, their lawyer asks for unlimited liability, waits for our lawyer between meetings, our lawyer rewrites the terms, waits for approval, signed day 21 after two chases. 2 hours of real work spread over 21 days. After: the agent drafts each contract on the company's standard terms and answers routine questions from the other side; the lawyer decides only the unusual terms. Yes to signed: 21 days to 3 days. Yeses that go cold: 1 in 10 to almost none. https://workweaver.ai/workflows/contracts/
14. Operations: stock runs out while the reorder waits. Before: a shift lead posts in the WhatsApp group, the message is lost among 60, cafes run out, staff buy cartons at the supermarket, the order is over USD 2,000 so it needs the owner, the supplier confirms for Thursday: 5 days with no oat milk. 15 minutes of real work spread over 6 days. After: the agent reads the tills and the stock counts and drafts the order before the shelf runs low; one tap on WhatsApp approves it. Days the cafes had no oat milk: 5 to 0. Oat drinks turned away: about 1,900 to 0. Emergency supermarket runs: 9 to 0. https://workweaver.ai/workflows/stock-outs/
15. Finance: 30-day invoices get paid on day 64. Before: a USD 48,000 invoice goes to the customer's general inbox, no reply, due on day 31, two more reminders unread, finance calls on day 41, the invoice is parked with no PO number and 200 cartons short, credit note waits on the warehouse, paid day 64 in the customer's next payment run. 35 minutes of real work spread over 64 days. After: the agent sends every invoice with the customer's PO number and finds the reason for a delay in the first week. Invoice to cash: 64 days to 30 days. When the short delivery came up: day 41 to day 2. Reminders before anyone talked: 3 unread to none needed. https://workweaver.ai/workflows/late-payments/
16. Accounting: closing the books takes 12 days. Before: bank statements for 4 hotels, 140 card payments with no receipt, emails to 38 managers, one hotel's bank off by USD 3,410, two days to find why, the laundry bill is not in so the accruals wait; books closed day 12 and the owner's meeting was day 10. About 30 hours of real work spread over 12 days of chasing. After: the agent asks for each receipt on WhatsApp the day the card is used, matches the bank lines and explains every item left over. Month-end close: 12 days to 4 days. Card payments with no receipt on day 1: 140 to 9. The owner's meeting on day 10: held without numbers to numbers ready on day 4. https://workweaver.ai/workflows/month-end-close/

## How customers start

Two ways to start. Self-serve: start alone with 7 free days from the first run, then the meter runs at USD 9 per agent-hour. Guided 90-day rollout: setup and training for WorkWeaver; agent work is billed at USD 9 an agent-hour. Book at https://workweaver.ai/ or start at https://workweaver.ai/signup.

Weeks 1-2: WorkWeaver maps where work waits: one process, step by step, with today's hours. Every step is sorted into delete (stop doing work nobody needs), rule (write it down once), agent (reading, checking, drafting, chasing) or human (judgment, trust, the final yes).

Weeks 3-6: the first workflow goes live inside the tools the team already uses: Slack, WhatsApp and email. The team spends about three hours a week learning to lead the agents.

Weeks 7-10: more workflows go live and the team leads them.

Weeks 11-13: the team's people lead it.

Included: operator training for 5 to 10 people so they lead the agents themselves; ready-made skills for jobs every company has (invoices, follow-ups, scheduling, reporting); and a next-bottleneck review at day 120 so the company knows where to go next.

## Questions, answered

Will it replace my staff? It replaces the chasing, copying and waiting, not the judgment. People move from human in the loop to human on the loop.

How is this different from a chatbot? A chatbot answers questions. WorkWeaver does the work: it joins meetings, takes calls, follows up and finishes jobs, and what it learns is shared with the whole company.

What does the guided rollout demand of my team? In the first two weeks, WorkWeaver maps where work waits. After that, the people who lead the agents spend about three hours a week. That is all it takes.

How soon will we see results? The first workflow runs inside the company's own tools between weeks three and six. By week thirteen, the company's people lead it on their own.

What does it cost to run? Agent work at USD 9 an agent-hour, priced by the minute, plus whatever the company's own providers charge for models, calls and messages. The first 7 days are free from the first run.

Which tools does it work with? The ones the company already uses: CRM, accounting, documents, calendar, project tools, Slack, Teams, WhatsApp, email and phone.

Where does company data live? In a workspace that is yours alone, hosted on Amazon Web Services. WorkWeaver is not locked to any one cloud. Export every skill and memory any time.

## Company record and contacts

BitFoundry Tech Private Limited. Brand: BitFoundry. Founded 2025. Headquarters: Coimbatore, Tamil Nadu, India. CIN: U62011TZ2025PTC035159. Member of the NVIDIA Inception Program. Product: WorkWeaver (https://workweaver.ai/). Pricing: USD 9 per agent-hour, no seats, no minimum. Runs on: Amazon Web Services; not locked to any one cloud. Founders: Aditya Koushik V, Co-founder and CEO (https://www.linkedin.com/in/aditya-koushik-v-33b35482/); Santhanakrishnan Dhandapani, Co-founder and CTO (https://www.linkedin.com/in/santhanakrishnan-dhandapani-156418184/, https://x.com/Santhanakrishna). Company: contact@bitfoundry.ai. Product: workweaver@bitfoundry.ai.

## For agents

Read this site as text: https://www.bitfoundry.ai/llms.txt and https://www.bitfoundry.ai/llms-full.txt. Page mirrors: https://www.bitfoundry.ai/index.txt, https://www.bitfoundry.ai/workweaver.txt, https://www.bitfoundry.ai/about.txt, https://www.bitfoundry.ai/pricing.txt. Full content is in the initial HTML; no step needs JavaScript. Sitemap: https://www.bitfoundry.ai/sitemap.xml. Agent map of WorkWeaver: https://workweaver.ai/llms.txt and https://workweaver.ai/llms-full.txt. MCP server https://mcp.workweaver.ai (remote MCP, OAuth): the WorkWeaver product itself, for agents acting for a signed-in WorkWeaver customer. A2A agent card https://api.workweaver.ai/.well-known/agent-card.json.