AgenticOrg

AI Virtual Employee Platform — LangGraph agents, 1000+ integrations (via Composio), native connectors and tools. Agents call real APIs (Jira, HubSpot, GitHub, GSTN, Tally, Banking AA) — not just generate text. Voice agents, RAG knowledge base, smart LLM routing, industry packs, PII redaction, browser RPA, CFO/CMO dashboards, ABM dashboard, NL Query (Cmd+K), multi-company support, scheduled reports, A/B testing, email drip engine, web push HITL, Python/TypeScript SDKs, MCP server, human-in-the-loop governance, no-code builder.

Live counts for agents, connectors, tools, and product version come from GET /api/v1/product-facts. Anywhere a count appears in the UI it is fetched from that endpoint — do not hand-edit numbers in docs or pages; they will drift.

Live: https://agenticorg.ai | App: https://app.agenticorg.ai | Playground: https://agenticorg.ai/playground

Latest Mainline Status (2026-06-30)

• Merchant self-service commerce config: /dashboard/commerce-runtime now lets tenant admins and merchant operators configure and edit merchant/store scoped source connectors, buyer channels, payment providers, public publishing, and Offline POS metadata. Runtime records are scoped by tenant, merchant, and seller agent; secret values stay in approved custody systems and config rows store only refs/redacted metadata.
• OACP Shopify runtime vertical: Seller Commerce Agent onboarding, encrypted merchant-scoped Shopify credential setup, read-only Shopify Admin GraphQL product/variant/price/inventory sync, Shopify webhook HMAC verification, Grantex C6Z authority request, 11-family OACP artifact cache intake, buyer Q&A from cache, protocol adapter payload generation, and bridge readiness endpoints are implemented.
• OACP runtime launch closure: docs/oacp/runtime-launch-closure-prd.md is the canonical launch checklist. OACP_LAUNCH_WRITE_EVIDENCE=true python scripts/oacp_runtime_launch_check.py emits redacted JSON/Markdown evidence with IDs, counts, statuses, public-catalog URLs, and exact external credential blockers.
• Buyer and channel bridges: Web/API answers, MCP seller tools, OpenAPI/A2A metadata, and WhatsApp/Telegram webhook routes share the same artifact-backed answer path and source/freshness labels. WhatsApp and Telegram require configured webhook secrets before accepting inbound messages.
• Mandate/payment/POS boundary: AgenticOrg verifies Pine Labs Plural/P3P mandate capability metadata, prepares non-executing purchase/mandate handoffs, and builds Offline POS Bridge packets with confirmation reconciliation. POS payment/receipt confirmations require server-side signed callback verification when OFFLINE_POS_WEBHOOK_SECRET is configured. It does not fake paid states, create orders, create checkout sessions, reserve stock, or execute provider/POS rails.
• Security hardening: production dependencies were trimmed, JWT runtime moved to PyJWT[crypto], python-jose/ecdsa are blocked by regression gates, Docker runtime no longer needs curl, and security CI fails closed on high-risk findings.
• Deployment hardening: production rollout uses the manual Cloud Run helper with split Cloud Run and Artifact Registry regions, image digest and commit metadata checks, migration-first rollout support, and explicit traffic modes.
• SDK launch coverage: Python SDK, TypeScript SDK, and MCP server now cover A2A/MCP discovery, commerce_sales_agent launch, connector listing, knowledge search, agent generation, workflow generation, workflow creation, workflow runs, and run-status polling in regression tests.

OACP Seller And Buyer Commerce Runtime

AgenticOrg owns the Seller Commerce Agent and buyer-agent runtime for OACP-backed commerce. It runs merchant self-service configuration, Shopify connector custody, read-only Shopify sync, buyer sessions, channel bridges, artifact cache, protocol adapter consumption, and provider-owned capability verification. Pine Labs Plural/P3P is the current verifier path; bank-owned and other rails remain adapter-gated.

Grantex owns OACP trust authority, protocol/policy governance, canonical signed/internal artifacts, artifact verification, and protocol adapter authority. Shopify and merchant systems remain the source of record. Pine Labs Plural/P3P owns mandate and payment rail execution.

flowchart LR
  merchant[Merchant operator] --> config[Merchant commerce config]
  config --> agentic[AgenticOrg Seller Commerce Agent]
  agentic --> shopify[Read-only Shopify sync]
  shopify --> grantex[Grantex OACP authority]
  grantex --> cache[AgenticOrg OACP cache]
  cache --> buyer[Buyer agent and channels]
  buyer --> provider[Pine Labs Plural/P3P capability check]

Read the canonical docs in docs/oacp/README.md, starting with docs/oacp/runtime-launch-closure-prd.md, docs/oacp/merchant-commerce-configuration.md, docs/oacp/end-user-flow.md, docs/oacp/truth-inventory.md, and docs/oacp/offline-pos-bridge.md. Older Commerce Sales Agent docs and reports remain historical/contextual where they describe earlier Grantex Commerce V1 tool paths.

What Is AgenticOrg?

AgenticOrg deploys AI virtual employees that automate enterprise back-office work. Each agent has a name, designation, specialization, and tailored instructions — like a real team member. They process invoices, run payroll, triage support tickets, launch campaigns, and reconcile bank statements — with human approval on every critical decision.

Key Numbers

What It Does

Agents That Act, Not Just Talk

Unlike AI chatbots that only generate text, AgenticOrg agents execute real actions:

You: "Production API returning 500 errors. CloudSQL connection pool exhausted."

Ops Commander (Aria Singh):
  → Gemini reasons about severity (636 tokens, 2.6s)
  → Creates Jira ticket KAN-5 via real API (1.1s)
  → Sets priority to Highest, assigns engineering team
  → Returns analysis + ticket link to you

Verified on production: agents have created 14 real Jira tickets, read HubSpot CRM data (contacts, deals, companies), and queried GitHub repo statistics — all through the LLM → tool_calls → connector → API pipeline.

Architecture

Landing Page (animations, interactive demo, blog, SEO, developer section)
    |
App Dashboard
    ├── CFO Dashboard (/dashboard/cfo) — Cash Runway, Burn Rate, DSO, DPO, AR/AP Aging
    ├── CMO Dashboard (/dashboard/cmo) — CAC, MQLs, SQLs, Pipeline, ROAS by Channel
    ├── NL Query (Cmd+K search bar + slide-out chat panel)
    ├── Multi-Company Switcher (CA firms managing multiple clients)
    ├── Agent Fleet, Workflows, Approvals, Audit, Sales Pipeline, Integrations
    └── Report Scheduler (create, manage, toggle, run-now scheduled reports)
    |
FastAPI Backend
    ├── Agent Registry (36 agents) → Smart LLM Router (RouteLLM)
    │       ↓ tool_calls                    ↑ synthesis
    │   Tool Gateway → 1000+ Integrations (Composio) + Native Connectors (see /api/v1/product-facts for counts)
    │       ├── Jira (11 tools) ← verified, creates real tickets
    │       ├── HubSpot (13 tools) ← verified, reads real CRM
    │       ├── GitHub (9 tools) ← verified, reads real repos
    │       ├── GSTN (8 tools) ← real Adaequare 2-step auth + DSC signing
    │       ├── Tally (7 tools) ← XML/TDL protocol + bridge for on-premise
    │       ├── Banking AA (5 tools) ← RBI-compliant consent flow
    │       ├── Bombora, G2, TrustRadius ← intent data (weighted 40/30/30)
    │       └── GA4, MoEngage, NetSuite, WordPress, Twitter/X, YouTube, Mailchimp...
    ├── NL Query Engine → Agent attribution → Chat history
    ├── Scheduled Report Engine → Celery beat → PDF/Excel → email/Slack/WhatsApp
    ├── LangGraph Runtime → GraphInterrupt HITL → Shadow Mode
    ├── Workflow Engine (20+ templates) → real agent execution → HITL Queue
    ├── NEXUS Orchestrator → Audit Logger
    ├── A2A Protocol → Agent Discovery → Cross-platform Tasks
    ├── Commerce Runtime → OACP-governed seller/buyer tools → fail-closed discovery and handoff
    ├── OACP Artifact Cache → durable public-safe refs → TTL/revocation/risk evaluation
    ├── OACP Cache Maintenance → keep/refresh/evict/purge/quarantine planner
    ├── RAGFlow Engine → Document ingestion → Semantic search → Agent retrieval
    ├── LiveKit + Pipecat → Voice agents → STT/TTS → SIP telephony
    ├── RouteLLM → Smart model routing → 3-tier cost optimization
    ├── Presidio → Pre-LLM PII redaction → Aadhaar/PAN/GSTIN scrubbing
    ├── Composio → 1000+ tool marketplace → OAuth bridge
    ├── MCP Server → governed agent + seller tools for Claude/Cursor/ChatGPT/VS Code
    ├── API Key Manager → ao_sk_ keys → SDK/CLI/MCP auth
    ├── SOP Parser → Upload SOPs → Deploy as agents
    └── Sales Agent → Gmail API → Email Sequences
    |
SDKs: Python (PyPI) | TypeScript (npm) | MCP Server | CLI
    |
PostgreSQL (Cloud SQL) + Redis + GCS + GCP Secret Manager

Agent Domains (6 Domains, 50+ Agents)

Pre-built agents across 6 domains + 4 industry packs. 1000+ integrations (via Composio) + native connectors with tools. All endpoints real. Live counts: GET /api/v1/product-facts.

Core Features

CFO Dashboard (/dashboard/cfo)

Role-specific KPI view for finance leaders. Real-time metrics powered by finance agents:

• Cash Runway — months of cash remaining at current burn rate
• Burn Rate — monthly cash outflow trend
• DSO / DPO — Days Sales Outstanding and Days Payable Outstanding
• AR/AP Aging — receivables and payables bucketed by 30/60/90/120+ days
• P&L Summary — revenue, COGS, gross margin, EBITDA
• Bank Balances — aggregated from all connected bank accounts via AA
• Tax Calendar — upcoming GST/TDS/advance tax deadlines with status

CMO Dashboard (/dashboard/cmo)

Role-specific KPI view for marketing leaders. Real-time metrics powered by marketing agents:

• CAC — Customer Acquisition Cost by channel
• MQLs / SQLs — Marketing and Sales Qualified Leads pipeline
• Pipeline Value — total opportunity value by stage
• ROAS by Channel — Return on Ad Spend for Google, Meta, LinkedIn
• Email Performance — open rate, CTR, unsubscribe, deliverability
• Brand Sentiment — positive/negative/neutral trend from social monitoring
• Content Performance — top pages by traffic, engagement, conversions

NL Query Interface

Natural language search across all your business data:

• Cmd+K Search Bar — global shortcut opens search from any page
• Slide-out Chat Panel — full conversational interface with agent attribution
• Agent Attribution — every answer shows which agent(s) provided the data
• Example Queries: "What's my cash position?", "Show me AP aging over 90 days", "How did Google Ads perform last week?", "What's our DSO this quarter?"

Multi-Company Support

Built for CA firms and holding companies managing multiple entities:

• Company Switcher — dropdown in the top nav to switch between client entities
• Isolated Data — each company has its own agents, connectors, workflows, and audit trail
• Cross-Company Reporting — consolidated views across all managed entities
• RBAC per Company — different roles and permissions for each entity

Scheduled Report Engine

Automated report generation and delivery:

• Celery Beat Scheduler — cron-based scheduling for any report
• Output Formats — PDF and Excel with branded templates
• Delivery Channels — email, Slack, and WhatsApp
• Report Scheduler UI — create, manage, toggle on/off, and run-now from the dashboard
• Pre-built Schedules — daily cash report, weekly P&L, monthly close package, weekly marketing digest

Workflow Templates (15 Pre-built)

Production-ready workflow templates that combine multiple agents:

Virtual Employee System

Each agent is a virtual employee with persona, specialization, and routing:

• Employee name (Priya, Arjun, Maya) — appears in audit logs and approvals
• Designation (Senior AP Analyst - Mumbai) — role clarity
• Specialization (Domestic invoices < 5L) — task routing
• Multiple agents per role — 3 AP Processors for different regions with smart routing

Organization Chart

Visual tree hierarchy per department. Each agent has an org_level field (e.g., Head → Manager → Analyst) and a parent reference, forming a reporting chain used for smart escalation — when an agent's confidence is below threshold, the task auto-escalates to its parent in the org tree. CSV bulk import lets you onboard entire departments in one upload.

No-Code Agent Creator

5-step wizard: Persona → Role → Prompt → Behavior → Deploy as Shadow. Or describe the employee you need in plain English — the system auto-generates the full agent configuration.

NL Workflow Builder

Describe business processes in plain English (e.g., "Automate invoice approval when amount > 5L") and the system generates the complete workflow definition. Preview before deploying.

Multi-Language Support

Platform available in English and Hindi. Language picker in header. Agents respond in the user's preferred language. Extensible to Tamil, Telugu, Kannada.

Content Safety

AI-generated content is checked for PII leakage, toxicity (via HuggingFace toxic-bert), and near-duplicate detection before delivery. Configurable thresholds per agent.

Prompt Template Library

27 production-tested templates with {{variable}} substitution. Built-in templates are read-only — clone to customize. Full prompt audit trail with edit history. Prompt lock on active agents.

Human-in-the-Loop (HITL)

Configurable confidence floors, trigger conditions, escalation chains, and timeout rules. Every HITL decision logged in audit trail. Agents cannot bypass their own gates. LangGraph-based HITL uses GraphInterrupt for pause/resume — the agent graph pauses at the approval node and resumes only after human decision.

RBAC

• CEO/Admin: Full access, all domains
• CFO: Finance agents only
• CHRO: HR agents only
• CMO: Marketing agents only
• COO: Operations agents only
• Auditor: Read-only audit log

Sales Agent

Automated lead qualification, personalized email outreach, follow-up sequences (Day 1/3/7/14), Gmail inbox monitoring, and pipeline dashboard. Demo request form → instant personalized response.

Native Enterprise Connectors — All Real API Endpoints

Finance (11): Oracle Fusion, SAP, Tally (XML/TDL + bridge), GSTN (Adaequare 2-step + DSC), Stripe, QuickBooks, Zoho Books, Banking AA (RBI consent), Income Tax India, Pine Labs (Plural), NetSuite HR (8): Darwinbox, Okta, Greenhouse, LinkedIn Talent, DocuSign, Keka, Zoom, EPFO Marketing (19): Salesforce, HubSpot, Google Ads, LinkedIn Ads, Meta Ads, Ahrefs, Mixpanel, Buffer, Brandwatch, GA4, MoEngage, WordPress, Twitter/X, YouTube, Mailchimp, Semrush, Bombora, G2, TrustRadius Ops (7): Jira, ServiceNow, Zendesk, PagerDuty, Confluence, Sanctions API, MCA Portal Comms (9): Slack, GitHub, Gmail, SendGrid, GCS, Google Calendar, Twilio, WhatsApp, LangSmith

Every connector uses real API endpoints from official documentation. Zero stubs. The Tally connector uses native XML/TDL protocol with a bridge agent for on-premise instances. Banking AA follows RBI-compliant consent flow. GSTN uses Adaequare's 2-step authentication with DSC signing for filing.

Tier 1: Marketing Automation (v3.2.0)

CA Firms -- Paid Add-on

INR 4,999/month per client | 14-day free trial

Purpose-built for Chartered Accountant firms managing multiple client companies from a single tenant.

Quick Start

Prerequisites

• Python 3.11+ (tested on 3.12, 3.13) | Node.js 20+ | Docker

Installation

AgenticOrg provides three requirements files for different deployment scenarios:

All versions in requirements.txt are pinned to exact production-tested versions (2026-04-06).

Local Development

git clone https://github.com/mishrasanjeev/agentic-org.git
cd agentic-org
cp .env.example .env  # Add your Gemini API key (free at aistudio.google.com)

# Install Python dependencies
pip install -r requirements.txt              # Core platform (required)
pip install -r requirements-v4.txt           # v4 features (optional)
pip install -r requirements-dev.txt          # Dev tools (optional)

# Start infrastructure (PostgreSQL + Redis)
docker compose up -d postgres redis

# Run API server
uvicorn api.main:app --host 0.0.0.0 --port 8000 --reload

# Run frontend (separate terminal)
cd ui && npm install && npm run dev

Optional Docker Services

docker compose --profile ragflow up -d       # Knowledge Base (RAGFlow)
docker compose --profile voice up -d         # Voice Agents (LiveKit)
docker compose --profile airgap up -d        # Local LLM (Ollama, CPU)
docker compose --profile airgap-gpu up -d    # Local LLM (vLLM, GPU)

Docker Compose

docker compose up -d
# API: http://localhost:8000 | UI: http://localhost:3000

Cloud Run Production

# Build/push and roll out origin/main with revision health checks.
bash scripts/deploy_cloud_run.sh --yes

# Deploy an already-built commit image and run migrations before traffic moves.
bash scripts/deploy_cloud_run.sh --sha <commit-sha> --skip-build --with-migrations --yes

# Stage revisions without moving production traffic.
bash scripts/deploy_cloud_run.sh --sha <commit-sha> --skip-build --traffic preserve --yes

Default production split: Cloud Run services run in asia-southeast1; Artifact Registry images live in asia-south1. Legacy GKE/Helm instructions remain in docs/deployment.md only for reference and non-default deployment shapes.

Demo Credentials

API Reference

Base URL: https://app.agenticorg.ai/api/v1

SDKs, CLI & MCP Server

pip install agenticorg is the canonical Python install path for both the SDK and the direct agenticorg CLI. Use the CLI from shell-capable assistants and developer environments such as Claude Code, Codex, Gemini CLI, VS Code tasks/terminals, CI jobs, and customer runbooks. Use the MCP server when the client supports Model Context Protocol and should call AgenticOrg as tools.

Python SDK

pip install agenticorg

from agenticorg import AgenticOrg

client = AgenticOrg(api_key="ao_sk_your_key_here")
result = client.agents.run("ap_processor", inputs={"invoice_id": "INV-001"})
agents = client.agents.list()
commerce = client.agents.run(
    "commerce_sales_agent",
    action="buyer_discovery_preview",
    inputs={"merchant_id": "merchant_demo", "query": "Show laptops under Rs 50000"},
)
draft_agent = client.agents.generate("Create a contract intelligence agent using Confluence and Jira")
kb = client.knowledge.search("vendor renewal policy", top_k=3)
workflow_draft = client.workflows.generate("Review vendor renewal risk using KB and Jira, then notify vendor_manager")
workflow = client.workflows.create(name="Renewal Risk Review", definition=workflow_draft["workflow"], domain="ops")
run = client.workflows.run(workflow["id"], payload={"vendor_id": "V-100"})
sop = client.sop.parse_text("When invoice > 5L, require CFO approval")
card = client.a2a.agent_card()

TypeScript SDK

npm install agenticorg-sdk

import { AgenticOrg } from "agenticorg-sdk";

const client = new AgenticOrg({ apiKey: "ao_sk_your_key_here" });
const result = await client.agents.run("ap_processor", { inputs: { invoice_id: "INV-001" } });
const commerce = await client.agents.run("commerce_sales_agent", {
  action: "buyer_discovery_preview",
  inputs: { merchant_id: "merchant_demo", query: "Show laptops under Rs 50000" },
});
const draftAgent = await client.agents.generate("Create a contract intelligence agent using Confluence and Jira");
const kb = await client.knowledge.search("vendor renewal policy", { topK: 3 });
const workflowDraft = await client.workflows.generate("Review vendor renewal risk using KB and Jira, then notify vendor_manager");
const agents = await client.agents.list();

MCP Server

Any MCP-compatible client (Claude Desktop, Cursor, ChatGPT, VS Code MCP clients) can use AgenticOrg agents and tools:

AGENTICORG_API_KEY=ao_sk_... npx agenticorg-mcp-server

Or add to your MCP client config:

{
  "mcpServers": {
    "agenticorg": {
      "command": "npx",
      "args": ["agenticorg-mcp-server"],
      "env": { "AGENTICORG_API_KEY": "ao_sk_..." }
    }
  }
}

CLI

pip install agenticorg

agenticorg agents list
agenticorg agents run ap_processor --input '{"invoice_id": "INV-001"}'
agenticorg agents run commerce_sales_agent --action buyer_discovery_preview --input '{"merchant_id":"merchant_demo"}'
agenticorg agents generate "Create a contract intelligence agent using Confluence and Jira"
agenticorg knowledge search "vendor renewal policy" --top-k 3
agenticorg workflows generate "Review vendor renewal risk using KB and Jira"
agenticorg workflows run wf-123 --input '{"vendor_id":"V-100"}'
agenticorg sop parse --text "When invoice > 5L, require CFO approval"
agenticorg a2a card
agenticorg mcp tools

The same root package also installs agenticorg-bridge for the on-prem Tally bridge used by CA-firm deployments.

See sdk/README.md and mcp-server/README.md for full documentation.

API Keys

API keys provide programmatic access for SDKs, CLI, and MCP integrations.

• Prefix: All keys use the ao_sk_ prefix (e.g., ao_sk_a1b2c3...)
• Generation: Admin users generate keys from Settings > API Keys in the app, or via POST /api/v1/org/api-keys
• Admin-only: API key endpoints require the agenticorg:admin scope
• Security: Keys are bcrypt-hashed at rest; the full key is shown only once at creation
• Revocation: Keys can be revoked instantly from Settings or via DELETE /api/v1/org/api-keys/{id}

Integration Protocols

A2A (Agent-to-Agent)

AgenticOrg implements Google's A2A protocol for cross-platform agent discovery and task execution:

• GET /a2a/agent-card — public agent discovery card (no auth required)
• POST /a2a/tasks — execute tasks via A2A protocol (JWT or Grantex auth)

MCP (Model Context Protocol)

MCP server exposing governed AgenticOrg agent tools, platform discovery, and read-only seller OACP artifact tools to compatible clients:

• GET /mcp/tools — list all available MCP tools (no auth required)
• POST /mcp/call — call an AgenticOrg MCP tool (JWT or Grantex auth)

Direct connector tools are not exposed as MCP tools. Use run_agent or agenticorg_<agent_type> tools so connector access remains inside the AgenticOrg runtime, scope checks, approval gates, and audit trail.

Grantex Authorization & Scope Enforcement

OAuth2-based authorization with manifest-based scope enforcement (Grantex SDK v0.3.3+):

• Manifest-based permissions: 53 pre-built tool manifests define exact permission levels (READ/WRITE/DELETE/ADMIN) for every connector tool — no keyword guessing
• grantex.enforce(): Offline JWT verification + manifest permission check in <1ms per tool call (JWKS cached after first call)
• Permission hierarchy: admin > delete > write > read — write scope covers read, admin covers everything
• LangGraph enforcement: validate_scopes graph node checks every tool call before execution
• ToolGateway enforcement: API-direct tool calls also use grantex.enforce() as primary enforcement
• POST /agents/{id}/delegate — set up Grantex delegation for agent-to-agent trust
• Scope hierarchy enforced: child agents cannot elevate parent scopes
• Token pool with automatic refresh at 50% TTL
• Custom manifests: place JSON/YAML files in GRANTEX_MANIFESTS_DIR for custom connectors
• Environment variables: GRANTEX_API_KEY (required), GRANTEX_BASE_URL (optional), GRANTEX_MANIFESTS_DIR (optional)

Environment Variables

COMPOSIO_API_KEY              — 1000+ tool integrations (optional)
AGENTICORG_LLM_ROUTING        — auto | tier1 | tier2 | tier3 | disabled
AGENTICORG_LLM_MODE           — cloud | local | auto
AGENTICORG_PII_REDACTION_MODE — before_llm | logs_only | disabled
RAGFLOW_API_URL               — RAGFlow instance URL for knowledge base
RAGFLOW_API_KEY               — RAGFlow API key
LIVEKIT_URL                   — LiveKit server URL for voice agents
LIVEKIT_API_KEY               — LiveKit API key
LIVEKIT_API_SECRET            — LiveKit API secret
STRIPE_SECRET_KEY             — Stripe billing key (hosted tier)
STRIPE_WEBHOOK_SECRET         — Stripe webhook signing secret

Knowledge Base (RAG)

Upload company documents (PDF, Word, Excel) and agents query them via semantic search. Powered by RAGFlow (Apache 2.0). Agents use knowledge_base_search tool automatically.

Voice Agents

Real-time voice AI with telephony support. Default STT: Whisper (local, Apache 2.0). Default TTS: Piper (local, MIT). Configurable SIP provider (Twilio/Vonage). Powered by LiveKit (Apache 2.0) + Pipecat (BSD-2).

Smart LLM Routing

Automatic multi-model routing via RouteLLM (Apache 2.0). Three tiers: Economy (Gemini Flash, free), Standard (Gemini Pro), Premium (Claude/GPT-4o). Reduces costs by 85% while maintaining accuracy. Air-gapped mode routes to Ollama/vLLM.

Pre-LLM PII Redaction

Microsoft Presidio (MIT) scrubs sensitive data before it reaches the LLM. Custom India recognizers: Aadhaar, PAN, GSTIN, UPI. Configurable: AGENTICORG_PII_REDACTION_MODE=before_llm|logs_only|disabled.

Industry Packs

Pre-built agent bundles: Healthcare (4 agents, HIPAA-aware), Legal (4 agents), Insurance (4 agents), Manufacturing (4 agents). One-click install via POST /packs/{name}/install.

Browser RPA

Playwright-based automation for legacy web portals without APIs. Pre-built scripts for Indian government portals (EPFO, MCA, Income Tax). Sandboxed Docker execution with screenshot audit trail.

Billing & Hosted Tier

Self-hosted is free forever. Hosted tier: Free (3 agents) / Pro ($49/mo) / Enterprise ($299/mo). India pricing: Free / Rs 999/mo / Rs 4999/mo. Stripe (global) + PineLabs Plural (India).

Explainable AI

Every agent decision includes a "Why?" panel with 3-5 plain-English bullet points, confidence bar, tools cited, and Flesch-Kincaid readability scoring. Non-technical users can understand why the agent approved or rejected an action.

Self-Improving Agents

Thumbs up/down feedback on every agent run. After 10+ feedback entries, the system auto-generates prompt amendments. Agents learn from corrections without manual prompt editing.

Multi-Agent Collaboration

Workflow collaboration step type runs 2+ agents in parallel with shared context. Aggregation strategies: merge (combine outputs), vote (majority wins), first_complete (fastest agent wins).

Customer Support Deflection

Pre-built support_deflector agent auto-resolves 60%+ of support tickets using FAQ matching + knowledge base (RAG) search. Tracks deflection rate metric on the dashboard.

Real-Time CDC (Change Data Capture)

Webhook receivers + polling-based CDC for connected systems. When data changes externally (e.g., new HubSpot deal), CDC events trigger workflows automatically. HMAC-SHA256 signature validation, fail-closed security.

Integration Workflow Page

The /integration-workflow page provides a visual guide for connecting external systems, with SDK/CLI quickstart examples and protocol documentation.

Testing

# Backend tests (1931+ total)
pytest tests/

# Frontend vitest (93 tests)
cd ui && npx vitest run

# Playwright E2E — 17 spec files, CI runs against production
npx playwright test

# Latest commerce/deploy/security guards
python -m pytest tests/unit/test_oacp_c6x4_durable_cache_repository.py tests/unit/test_oacp_c6x5_cache_maintenance.py tests/unit/test_deploy_cloud_run_script.py tests/regression/test_security_audit_20260613_dependency_gates.py -q

# Production connector test — real Jira/HubSpot/GitHub API calls
python tests/test_production_connectors.py

Project Structure

agenticorg/
├── api/v1/                 # FastAPI endpoints (agents, auth, sales, templates, connectors, api-keys, sop, a2a, mcp)
├── core/
│   ├── agents/             # 50+ agent types + 27 prompt templates
│   │   ├── prompts/        # Production system prompts
│   │   ├── registry.py     # Agent registry (built-in + custom type fallback)
│   │   └── base.py         # BaseAgent: LLM reasoning → tool calling → HITL
│   ├── langgraph/          # LangGraph agent graph, runner, Grantex auth, LLM factory
│   ├── orchestrator/       # NEXUS: task routing, smart routing, state machine
│   ├── llm/                # LLM router (Gemini primary, Claude/GPT-4o fallback)
│   ├── commerce/           # Commerce/OACP helpers, cache evaluator, maintenance planner
│   ├── models/             # SQLAlchemy ORM (agents, workflows, HITL, leads, templates, api_keys, OACP cache)
│   ├── tool_gateway/       # Scope enforcement, rate limiting, PII masking, audit
│   ├── gmail_agent.py      # Gmail API integration (inbox monitor, send replies)
│   └── email.py            # SMTP email sending
├── connectors/             # Native enterprise connectors (counts: /api/v1/product-facts)
│   ├── finance/            # Oracle, SAP, GSTN, Stripe, Tally, Banking AA, NetSuite... (11)
│   ├── hr/                 # Darwinbox, Okta, Greenhouse, EPFO, Zoom... (8)
│   ├── marketing/          # Salesforce, HubSpot, Google Ads, GA4, MoEngage, Bombora, G2, TrustRadius... (19)
│   ├── ops/                # Jira, Zendesk, ServiceNow, PagerDuty... (7)
│   ├── comms/              # Slack, GitHub, Gmail, SendGrid, GCS, Twilio... (9)
│   └── framework/          # BaseConnector, auth adapters, circuit breaker
├── auth/                   # Grantex middleware, registration
├── workflows/              # Workflow engine, triggers, conditions
├── sdk/                    # Python SDK (pip install agenticorg) + CLI
├── sdk-ts/                 # TypeScript SDK (npm i agenticorg-sdk)
├── mcp-server/             # MCP Server (npx agenticorg-mcp-server)
├── ui/src/
│   ├── pages/              # Landing, Dashboard, Agents, Sales, Blog, Resources, Integrations, IntegrationWorkflow, ConnectorCreate, ConnectorDetail, Settings
│   ├── components/         # AgentCard, ActivityTicker, InteractiveDemo, SocialProof
│   ├── pages/blog/         # 8 SEO blog articles
│   └── pages/resources/    # 26 SEO content pages across 7 topic clusters
├── tests/
│   ├── unit/               # Unit tests
│   ├── security/           # Security tests
│   ├── connector_harness/  # Connector tests (every native connector × exposed tools)
│   ├── regression/         # Regression tests
│   ├── integration/        # Integration tests
│   ├── synthetic_data/     # Invoice, resume, contract test data
│   └── e2e/                # Playwright browser tests
├── ui/e2e/                 # 17 Playwright E2E spec files
├── ui/tests/               # 93 vitest component tests
├── migrations/             # PostgreSQL DDL files
├── helm/                   # Kubernetes Helm charts
├── docs/                   # PRD, architecture, QA test plan
└── scripts/                # Seed data, Cloud Run deploy helper, migration helper

SEO & AI Discoverability

Pricing

Self-hosted is free forever — all agents, all connectors, unlimited tasks.

Payments: Stripe (global) + PineLabs Plural (India: NEFT/RTGS/IMPS/UPI).

India-First Connectors

Built for Indian enterprise — not retrofitted:

• GSTN — e-Invoice IRN, GSTR-1/2A/3B/9 filing, ITC reconciliation
• EPFO — ECR filing, UAN verification, passbook download
• Income Tax India — TDS 26Q/24Q, Form 16A, 26AS credit check
• Darwinbox — HRMS (payroll, attendance, performance, onboarding)
• Tally Prime — Native XML/TDL protocol + bridge agent for on-premise tunneling (vouchers, trial balance, GST reports)
• Banking AA — RBI-compliant Account Aggregator consent flow (read-only: statements, balances, transactions)

Security & Compliance

Documentation

Contributing

Fork → branch → test → PR. See CONTRIBUTING.md.

• Python: ruff check . && mypy --ignore-missing-imports .
• TypeScript: npx tsc --noEmit
• Tests: pytest tests/unit/ tests/connector_harness/

License

Apache License 2.0 — free for commercial use. See LICENSE.

Built by AgenticOrg | Bengaluru, India