Documentation — AxonBOS

Overview

AxonBOS (Business Operating System) is an AI-powered integration and intelligence platform that connects ANY business application — not just Tally. Think of it as the nervous system of your business: it receives signals from payment gateways, bank feeds, e-commerce platforms, ERPs, custom SQL databases, and more, processes them intelligently, and delivers clean, verified actions.

Why AxonBOS?

Global iPaaS platforms (Zapier, Make, Workato) don't understand GST, charge 10-100x more, and can't talk to Indian systems natively. AxonBOS is built ground-up for businesses that need real integration — with a self-evolving connector framework that gets smarter with every user.

Key Capabilities

Self-Evolving Connectors — An AI discovery engine that auto-discovers and auto-configures connectors across 5 regions and 15+ business domains
AI Chat — Query your business data in plain English, Hindi, Kannada, Tamil, Telugu, or 10+ more languages
Multi-System Support — Tally, SQL Server, PostgreSQL, REST APIs, Excel, File Systems — any data source
Human-in-the-Loop — Every write operation requires your explicit approval
Self-Service Builder — Define your own connectors in natural language. No code needed.
Multi-Company — Each company gets its own subsite, dashboard, and role-based access
Pay As You Go — Start at ₹499. No annual contracts. Wallet-based recharge model.

Quick Start

Get started in under 15 minutes with the demo or your own data.

Option 1: Try the Demo

Visit demo.axonbos.com to explore 6 pre-built demo companies with sample data — manufacturing (steel), building supplies (NZ), IT services (CRM), food & beverage, facilities management, and real estate.

Option 2: Connect Your Own System

          Steps
          1. Sign up at app.axonbos.com

Go to Connectors → Add New

Pick your data source (Tally, SQL, API, etc.)

Provide credentials → AI auto-discovers schema

Review generated metadata → confirm

Start querying via AI Chat → Done!

Beta Status

AxonBOS beta go-live is March 15, 2026. The demo at demo.axonbos.com is live with sample data. Production connectors are being validated.

Architecture

AxonBOS uses a declarative, Markdown-native architecture. All business data, rules, and configurations are stored as human-readable Markdown in PostgreSQL — making the system both AI-parseable and user-editable.

          Core Pipeline
          User Prompt (any language)

  → AI parses intent

  → Fetch connector metadata (Markdown) from PostgreSQL

  → Match against business rules

  → Fire attached plugins (SMS, email, PDF, Excel)

  → Execute on target system (SQL, Tally, API, File)

Three Pillars

Axon Connect — Communication layer (Rocket.Chat based). Voice-first data entry. Supports Hindi and regional languages.
Axon Core — FastAPI backend with adaptive SQL. PostgreSQL + pgvector. Multi-tenant via company_id. AI models: Claude Sonnet for reasoning, Gemini Flash for high-volume operations.
Axon Agent — Desktop service for file monitoring, Excel sync, Tally integration. Bridges legacy tools into the modern stack. Works offline.

Database Architecture

The system runs on 33 PostgreSQL tables organized into 11 functional groups: Core Platform (7), Security (2), Learning Engine (4), Flywheel Accelerators (7), Orchestration (2), UI/Subsites (2), Operations (4), Billing (1), Versioning (1), AI Discovery (2), and Event Listeners (2).

Self-Evolving Connector Framework

AxonBOS doesn't have 222 hardcoded connectors. It has a framework that auto-discovers and auto-configures connectors for ANY system. The first connection uses high-reasoning AI (Claude Sonnet/Opus). Every subsequent connection for the same system type is instant — pulled from the shared schema registry.

          Connector Pipeline
          Prompt → PostgreSQL Markdown → Rules → Target System

Target Systems: SQL, Excel, File System, API, Tally XML

Plugins (attach to any connector):

  SMS | Email | PDF Generation | Excel Download | WhatsApp (future)

Building a new connector = writing a Markdown doc + rule definitions. Not a new codebase. Target: 90 minutes for the first connector of a type. 20 minutes once patterns are established.

AI Discovery Engine

When building a connector for a NEW external system for the first time, AxonBOS doesn't do template lookup — it reasons like a senior AI engineer. Full research, hypothesis, code generation, sandbox execution, validation, self-correction, and crystallization.

The 7-Step Reasoning Loop

Research — Search for API docs, OpenAPI/WSDL specs, community forums
Hypothesize — Generate candidate request (XML for Tally, REST for SAP, SQL for databases)
Code — Write Python test scripts to connect, send requests, parse responses
Execute — Run in isolated Docker sandbox (30s timeout, network locked to target only)
Observe — Classify result: SUCCESS / AUTH_FAILURE / SCHEMA_MISMATCH / TIMEOUT / ERROR
Correct — If failed: search docs for fix, modify hypothesis, re-execute (max 5 loops)
Crystallize — On success: generate metadata, store in schema registry with pgvector embedding

"Watch Claude Think"

The entire discovery process streams to you in real-time via WebSocket. You see exactly what the AI engineer is thinking and doing — researching docs, writing code, analyzing responses, and correcting errors. This builds trust and lets you intervene if needed.

Cost Model

First-time discovery: ~10-20 credits (high-reasoning AI model, multiple iterations)
Subsequent connections (template exists): ~1 credit (cached lookup + quick validation)
Edge case corrections: ~1-5 credits depending on complexity

Self-Service Connector Builder

Users define their OWN connectors. No code. No developer. No ticket to the AxonBOS team.

User Flow

Define Connector — Describe your data source in natural language (even in vernacular). System auto-generates Markdown metadata.
Define Rules — Set business rules in plain language: "Send SMS when invoice is overdue by 7 days."
Provide Credentials — Enter connection details. Secure credential vault stores them (AES-256 encrypted).
Select Plugins — Pick from: SMS, Email, PDF generation, Excel download, WhatsApp (future).
Test & Go Live — System validates, runs test rule, confirms plugin triggers. Live in minutes.

Every user who creates a connector makes the platform smarter for the next user. This is the data flywheel — connector patterns, language embeddings, and rule templates compound automatically.

Multi-Company Support

Every company gets a dedicated subsite URL with a personalized dashboard. No shared generic UI.

          URL Structure
          axonbos.com/{company-slug}/dashboard

axonbos.com/{company-slug}/invoices

axonbos.com/{company-slug}/inventory

axonbos.com/{company-slug}/reports

axonbos.com/{company-slug}/settings

axonbos.com/{company-slug}/flows

axonbos.com/{company-slug}/chat

Dashboards are auto-generated from connector metadata. KPI cards, charts, and lists are role-aware — the accountant sees financial data, the warehouse manager sees inventory. Same URL, different views based on role.

Tally Integration

TallyPrime is the flagship connector — deep XML integration with ALTERID-based change detection, voucher creation, and ledger mapping.

          Tally Sync Flow
          Tally XML API (Port 9000)

  → ENVELOPE/HEADER/BODY structure

  → ALTERID-based change detection (poll every 5 min)

  → Smart polling: 5 min during business hours, 30 min at night

  → Only fetches changes, not full dataset

Capabilities

Export ledgers, vouchers, stock items, groups
Create vouchers with proper GST treatment and narrations
Multi-branch consolidation
Auto-reconciliation with payment gateways and bank feeds

SKM Steels — Live Customer

SKM Steels (Nagpur, manufacturing) is a live customer using the Safal ERP connector with SQL Server integration. 18 whitelisted tables, 166 forms, 1317 fields — all auto-discovered.

Payment Gateways

Auto-reconcile every payment. Settlement reports flow directly into your accounting system with proper GST treatment.

Razorpay — Webhook-based. payment.captured events auto-update invoice status.
Cashfree — Settlement and payout reconciliation.
PayU — Transaction matching and ledger posting.
PhonePe Business — UPI payment capture and reconciliation.

Bank Feeds

Bank statements auto-imported and matched against entries. No more manual CSV downloads.

HDFC Bank, ICICI Bank, SBI, Axis Bank
Account Aggregator (Setu) integration for consent-based bank data
PDF bank statement parsing via AI

E-Commerce

Orders, returns, and settlements from every marketplace — mapped to your books with correct GST and inventory adjustments.

Shopify — Orders, refunds, inventory sync
WooCommerce — WordPress-based store integration
Amazon FBA — FBA settlements, returns, inventory
Flipkart — Marketplace reconciliation

ERP & Custom SQL Connectors

AxonBOS connects to ANY SQL database or REST API — not just Tally. The AI Discovery Engine auto-discovers schema, maps fields, and generates the connector metadata.

Supported Data Sources

SQL Server — Direct connection with Change Data Capture (CDC) support
PostgreSQL — LISTEN/NOTIFY for real-time change detection
MySQL — Binlog streaming via Debezium
REST APIs — Zoho, Salesforce, SAP B1, any OpenAPI-compliant system
Excel/CSV — File watch on SFTP/S3/local directories
Custom ERP — Any system with a database or API endpoint

Safal ERP Example

The Safal connector (SKM Steels) demonstrates full SQL Server integration — 18 tables, schema auto-discovery, fuzzy table resolution, and 10 read tools + 10 write tools with human-in-the-loop approval.

SMS Intelligence (AxonML)

Parses payment SMS via a 3-level AI pipeline. Every UPI notification becomes a verified entry automatically.

UPI payment alerts → auto-match to customer/invoice
Bank SMS parsing → transaction extraction
Now a standalone plugin — attaches to any connector
Getting its own Dashboard UI + subsite URL

Talk to Your Business — AI Chat

The AI Chat lets you interact with your business data using natural language. Ask questions, generate reports, create entries — all through conversation. Works across ALL connected systems, not just Tally.

Query Examples

Ask anything about your business in plain language — in English or any Indian language:

          "How many invoices did we generate this month?"

"Show top 5 customers by outstanding balance"

"What's our GST liability for Q3?"

"List all payments received from Sharma Enterprises"

"Compare this month's revenue vs last month"

"Show stock below reorder level"

"aaj ki bikri kitni hui?" (Hindi)

"nanna invoice yavaga baratte?" (Kannada)

Write Operations

When you ask the AI to create, update, or modify data, it triggers a write operation with strict approval workflow:

AI generates the proposed action (e.g., a new journal entry)
You see a preview with full details — amounts, ledgers, GST treatment
You explicitly approve or reject
If approved, the entry is created with a full audit trail

Multi-Language Support

AxonBOS processes business commands in Hindi, Kannada, Tamil, Telugu, Marathi, and 10+ more Indian languages — not translations, but native understanding of how Indian business owners actually talk. A Nagpur steel trader describes inventory differently than a Chennai textile manufacturer. The language intelligence engine learns from every interaction.

Roles & Permissions (Natural Language RBAC)

Roles are defined by the company owner in plain English (or any vernacular). The AI converts natural language to structured Markdown and enforces restrictions automatically.

          Example
          Owner says: "Priya is my accountant, she can see invoices

but not salaries, she can't delete anything"

AI generates Role Markdown:

## Role: Accountant — Priya

- Can view all invoices and payments

- Can create new invoices

- Cannot delete any records

- Cannot view salary or employee financial data

- Can export to Excel and PDF

- Cannot modify GST settings

When Priya logs in, her role Markdown is injected into the AI system prompt BEFORE any data access. The AI naturally refuses restricted operations.

Scheduled Flows & Automation

Users define scheduled automations via chat prompts. Powered by n8n as the execution engine.

          Example
          User: "Every Monday 9 AM, pull last week's unpaid invoices,

generate PDF, email to me, SMS alert if total > ₹1L"

→ AI converts to n8n workflow

→ User tests manually first

→ Once tested, flow is scheduled

→ Every execution logged and monitored

Flows support: scheduled (cron), event-driven (rule triggers), and cross-connector (Connector A triggers Connector B). All flows must pass manual test before activation.

Always-On Event Listeners

AxonBOS doesn't just poll. It listens. Five listener types handle any system that pushes data:

Webhook Receiver — HMAC-verified webhooks from Razorpay, Shopify, Stripe, etc.
Polling Sync — For Tally and legacy DBs. ALTERID-based change detection.
Database CDC — SQL Server Change Tracking, PostgreSQL logical replication, MySQL Debezium.
File Watch — Monitor SFTP/S3/local directories for new files.
Email Parse — IMAP mailbox watch. Parse PO emails, supplier invoices.

Company Dashboard

Every company gets an auto-generated dashboard. KPI cards, charts, and lists are generated dynamically from connector metadata — no hardcoded views.

Revenue trends, invoice aging, stock levels, cash flow charts
Recent invoices, outstanding payments, low stock alerts
User-configurable via chat: "Show me daily sales graph and top 5 overdue invoices on my dashboard"
Role-aware: accountant sees finances, warehouse manager sees inventory
Edit any record via prompt window (not forms) — supports vernacular input

API Reference

AxonBOS exposes a RESTful API for programmatic access. The base URL for all API calls is:

https://chat-api.axonbos.com/api/v1

API Endpoints

Method	Endpoint	Description
GET	`/`	Health check
GET	`/health`	Detailed service health
GET	`/api/v1/connectors`	List available connectors
GET	`/api/v1/chat/demos`	List demo companies
POST	`/api/v1/chat`	Send message (synchronous)
POST	`/api/v1/chat/stream`	Send message (SSE streaming)
POST	`/api/v1/chat/actions/{id}/approve`	Approve/reject write action

Chat Request

          POST /api/v1/chat
          {

  "message": "Show me all invoices from February",

  "conversation_id": null,

  "connector_hint": "safal",

  "demo_id": "india"

}

Chat Response

          Response
          {

  "conversation_id": "abc-123-def-456",

  "message": "Found 5 invoices from February...",

  "action": null,

  "metadata": {

    "intent": "query",

    "connector": "safal",

    "iterations": 2,

    "model_used": "gemini-3-flash-preview",

    "cost_credits": 0.3

  }

}