What we solve

Six kinds of problem, one way of building.

Every build is tailored to the business. Here is what we typically make, and what it replaces.

Custom workflows

Processes that eat hours every week: copying between systems, chasing approvals, preparing the same report again.

What we build

An agent-run workflow shaped to your process, with approval steps where a person must decide and a log of everything it does.

For example

Month-end close, quote to order, onboarding a new supplier.

Analysis agents

The answers are in your ledgers, sheets and reports, but getting at them takes someone a day.

What we build

An agent your team can ask in plain language. It queries your data read-only, shows its working and names its sources.

For example

Who owes us over ₹2 lakh, and is it getting worse?

Chatbots & assistants

Customers, dealers and staff asking the same questions all day.

What we build

Assistants on WhatsApp, your website or internal chat, grounded in your own documents and systems, handing over to a person when it matters.

For example

Order status for dealers, a shop-floor SOP assistant, an internal IT and HR helpdesk.

Document & data intake

Paper, PDFs, emails and photos that someone types in by hand.

What we build

Agents that read them, check them against your records, post what is certain and escalate what is not.

For example

Supplier invoices, delivery challans, bank statements.

Integrations agents can use

Systems that do not talk to each other, and AI that cannot safely reach them.

What we build

Connectors and MCP servers that give agents narrow, permissioned, logged access to Tally, Zoho, SAP Business One, SQL databases, WhatsApp, email and spreadsheets.

For example

A read-only MCP server over your ledger.

Custom internal products

A tool your team keeps asking for that no off-the-shelf software fits.

What we build

A tailored web app or portal, built agent-native: the routine work happens inside it, run by agents, with people approving.

For example

A supplier onboarding portal, a dealer ordering portal, an approvals dashboard.

What we build with

The agentic toolkit, chosen per problem.

We use whichever of these fits. We prefer direct integrations and MCP wherever they exist, because they are faster and more reliable than an agent operating a screen.

MCP servers

An open standard for connecting AI to your systems. We expose your data and actions to agents through narrow, permissioned tools.

Agent skills

Packaged instructions, templates and scripts that teach an agent your way of doing a repeatable job, loaded only when needed.

Tool calling

The agent chooses from a list of approved actions, such as "look up an order". Your own software runs them, so the AI never touches your systems directly.

Knowledge retrieval

The agent searches your own documents first and answers from what it finds, with the source, instead of guessing.

Approval workflows

The agent does the routine work and pauses for a person to approve anything risky, then carries on.

Chat & messaging channels

WhatsApp, web chat and internal chat, with a clear route to a person. WhatsApp assistants are built for your business and your customers, as Meta's business terms require (not a general-purpose AI). They say they are an AI, only message people who opted in, and hand over to a person on request.

Browser & computer-use agents

For software with no API, an agent can operate the screen like a staff member. Slower and less certain than an integration, so we pilot it first.

Evaluation & guardrails

A test set of your real cases that every change is scored against, plus limits on what the agent may do.

Audit logging

A record of every question, action, decision and approval, so you can see what happened and why.

Where we would start

Starting points for Indian MSMEs.

Businesses of the same type share the same headaches. These are the places we would begin, tailored to how your business actually runs. They are starting points, not boxed products.

Distribution & wholesale

Owner-run distributors of FMCG, pharma and hardware, with field salesmen and thin margins.

Where the time goes
  • Orders arrive on WhatsApp, voice notes and calls, then get re-keyed
  • Chasing outstanding payments by phone
  • Scheme and discount calculations
  • Batch and expiry tracking
  • GST reconciliation and e-way bills

Typically Marg, Busy or Tally, with Excel and WhatsApp.

Where we would start
WhatsApp order to invoice

Reads text, voice and photo orders, matches products and customers, drafts the entry, and flags credit-limit breaches for approval.

Collections assistant

Ranks overdue parties from the ledger, sends polite reminders with a payment link, and records promises to pay.

Expiry and slow-mover alerts

Warns about near-expiry batches and suggests returns or discounts.

Not on the list? Most of our work isn't. Tell us the problem.