Skip to content

Autonomous AI bookkeeping for SMBs — 110 AI agents handle invoices, reconciliation, and financial reporting so you never touch a spreadsheet again.

License

Notifications You must be signed in to change notification settings

Alexi5000/ClawKeeper

Repository files navigation

ClawKeeper Logo

ClawKeeper

Your AI Bookkeeper That Never Sleeps

110 autonomous AI agents working 24/7 so you never touch a spreadsheet again.

License: MIT Built with Bun TypeScript React DeepSeek AI

Quick Start · Architecture · Features · API Docs · Contributing


ClawKeeper Dashboard

The Problem

Small and medium businesses spend 120+ hours per year on bookkeeping. Manual invoice processing, bank reconciliation, and financial reporting drain time, money, and sanity. Hiring a full-time bookkeeper costs $45K–$65K/year. Outsourcing to a firm? $500–$2,500/month.

The Solution

ClawKeeper deploys a full AI accounting department — a CEO agent that orchestrates 9 team leads and 100 specialized workers — to handle your entire bookkeeping operation autonomously. One natural language command. 110 agents. Zero spreadsheets.

"Generate monthly P&L report and reconcile all accounts"

That's it. ClawKeeper's CEO agent decomposes the task, delegates to the right teams, and delivers results in real-time.


Why ClawKeeper?

Feature Traditional ClawKeeper
Invoice Processing Manual entry, 5–10 min each AI OCR + auto-categorization in seconds
Bank Reconciliation Hours of matching transactions Automatic matching with discrepancy detection
Financial Reports Days to compile Generated on-demand, any time
Cost $45K–$65K/year (in-house) Self-hosted, pay only for AI inference
Availability Business hours 24/7/365
Multi-Tenant One client at a time Unlimited tenants, full data isolation

Features

  • One-Prompt Deployment — Deploy 110 AI agents with a single natural language command
  • Autonomous Invoice Processing — AI-powered OCR via Google Document AI, validation, and categorization
  • Bank Reconciliation — Automatic transaction matching and discrepancy detection via Plaid
  • Financial Reporting — P&L, Balance Sheet, Cash Flow, AP/AR Aging, and custom reports
  • Multi-Tenant Architecture — Complete tenant isolation with Row-Level Security and RBAC
  • 110-Agent Hierarchy — CEO + 9 orchestrators + 100 specialized workers
  • DeepSeek AI Integration — Cost-efficient AI reasoning (10–100x cheaper than GPT-4)
  • Real-Time Execution Streaming — Watch agents collaborate live via WebSocket + SSE
  • Production Security — Rate limiting, circuit breakers, PII detection, audit trails
  • Modern Dashboard — React + Vite + Tailwind + Shadcn with Command Center UI

Architecture

ClawKeeper Architecture

ClawKeeper follows a hierarchical multi-agent architecture inspired by how real accounting departments operate:

ClawKeeper CEO (Top-Level Orchestrator)
│
├── CFO Agent ──────────────── 8 workers  (Strategic planning, forecasting, budgeting)
├── Accounts Payable Lead ──── 15 workers (Invoice processing, payments, expense tracking)
├── Accounts Receivable Lead ─ 15 workers (Customer invoicing, collections, revenue tracking)
├── Reconciliation Lead ────── 12 workers (Bank matching, discrepancy resolution)
├── Compliance Lead ─────────── 10 workers (Tax compliance, audit preparation)
├── Reporting Lead ──────────── 12 workers (Financial reports, dashboards, analytics)
├── Integration Lead ────────── 12 workers (Plaid, Stripe, QuickBooks, Xero sync)
├── Data/ETL Lead ───────────── 10 workers (Data import, transformation, validation)
└── Support Lead ────────────── 6 workers  (User assistance, error recovery)

Tech Stack

Layer Technology
Runtime Bun
Language TypeScript (strict mode)
API Server Hono
Dashboard React 18 + Vite + Tailwind CSS + Shadcn/UI
Database PostgreSQL with RLS + RBAC
AI Engine DeepSeek AI
OCR Google Document AI
Banking Plaid
Payments Stripe
Accounting Sync QuickBooks, Xero
Real-Time WebSocket + Server-Sent Events

Quick Start

Detailed setup instructions: See STARTUP.md for troubleshooting and advanced configuration.

Prerequisites

1. Clone and Install

git clone https://github.com/Alexi5000/ClawKeeper.git
cd ClawKeeper
bun install

2. Configure Environment

cp .env.example .env

Set your DEEPSEEK_API_KEY in .env (and optionally Plaid, Stripe, QuickBooks, Xero keys).

3. Setup Database

# One-command setup (schema + RLS + RBAC + seed data)
bun run setup:full

Or manually:

createdb clawkeeper
bun run db:setup
bun run setup:demo

4. Start Services

# Terminal 1: API Server (default port 9100)
bun run dev

# Terminal 2: Dashboard (default port 3000)
bun run dashboard:dev

5. Launch the Command Center

Open http://localhost:3000 and log in:

Field Value
Email admin@demo.com
Password password123

Navigate to Command Center, type a command like "Generate monthly P&L report and reconcile all accounts", and watch 110 agents go to work.


Skills

ClawKeeper ships with 8 production-ready skills that agents use to perform specialized tasks:

Skill Description
invoice-processor OCR extraction, validation, and categorization of invoices
bank-reconciliation Automatic transaction matching and discrepancy detection
financial-reporting P&L, Balance Sheet, Cash Flow, and custom report generation
payment-gateway Payment processing and disbursement via Stripe
compliance-checker Tax compliance validation and audit trail generation
document-parser Multi-format document parsing (PDF, images, spreadsheets)
data-sync Bidirectional sync with QuickBooks and Xero
audit-trail Immutable logging of all financial actions

API Reference

Authentication

POST /api/auth/login          # User authentication
POST /api/auth/register       # New user registration

Invoices

GET    /api/invoices              # List invoices
POST   /api/invoices/upload       # Upload invoice for AI processing
POST   /api/invoices/:id/approve  # Approve invoice
POST   /api/invoices/:id/pay      # Mark as paid

Reports

GET /api/reports/:type

Supported types: profit_loss, balance_sheet, cash_flow, ap_aging, ar_aging

Reconciliation

POST /api/reconciliation/start       # Start reconciliation task
GET  /api/reconciliation/:id/status  # Check reconciliation status

Real-Time

WS /ws    # WebSocket for live agent execution updates

Full API documentation: docs/API.md


Security

ClawKeeper is built for production financial workloads with enterprise-grade security:

  • Row-Level Security (RLS) — Tenant data isolation enforced at the database level
  • Role-Based Access Control — 4 roles: super_admin, tenant_admin, accountant, viewer
  • Immutable Audit Trail — Every financial action is logged and tamper-proof
  • Rate Limiting — Per-tenant and per-endpoint rate limits
  • Circuit Breaker — Automatic protection for external API failures
  • Input Validation — Zod schemas enforce strict typing on all API inputs
  • PII Detection — Prevents sensitive data from leaking to LLM providers

Full security documentation: SECURITY.md


Project Structure

ClawKeeper/
├── agents/                # 110 agent definitions (CEO + orchestrators + workers)
│   ├── clawkeeper/       # CEO agent
│   ├── orchestrators/    # 9 team lead agents
│   └── workers/          # 100 specialized worker agents
├── skills/                # 8 production-ready skills
├── src/
│   ├── api/              # Hono API server
│   ├── agents/           # Agent runtime implementations
│   ├── core/             # Types, orchestrator engine, scheduler
│   ├── integrations/     # Plaid, Stripe, QuickBooks, Xero, Document AI
│   ├── memory/           # Agent memory and context system
│   ├── guardrails/       # Security, validation, PII detection
│   └── db/               # Database queries and migrations
├── dashboard/             # React admin dashboard (Vite + Tailwind + Shadcn)
├── db/                    # SQL schema, RLS policies, RBAC, seed data
├── docs/                  # Extended documentation
├── config/                # Configuration files
└── scripts/               # Deployment and utility scripts

Documentation

Document Description
STARTUP.md Detailed startup guide with troubleshooting
Architecture System design and agent hierarchy
API Reference Complete API documentation
Deployment Guide Production deployment instructions
Multi-Tenancy Tenant isolation and RBAC configuration
Security Security model and best practices
Contributing Contribution guidelines

Development

Adding a New Agent

  1. Create the agent definition: agents/<category>/<name>/AGENT.md
  2. Implement the runtime: src/agents/<name>.ts
  3. Register in src/agents/index.ts
  4. Deploy: ./scripts/deploy.sh

Adding a New Skill

  1. Create the skill definition: skills/<name>/SKILL.md
  2. Update the skills index: SKILLS.md
  3. Deploy: ./scripts/deploy.sh

Testing

bun test              # Run all tests
bun run typecheck     # TypeScript type checking
bun run lint          # ESLint

Roadmap

  • Multi-currency support
  • AI-powered cash flow forecasting
  • Automated tax filing preparation
  • Mobile companion app
  • Slack/Teams integration for notifications
  • Custom agent builder UI

License

MIT — see LICENSE for details.


Built by Alex Cinovoj · Powered by TechTide AI

Stop doing bookkeeping. Start building your business.

About

Autonomous AI bookkeeping for SMBs — 110 AI agents handle invoices, reconciliation, and financial reporting so you never touch a spreadsheet again.

Topics

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Packages

No packages published