DEV Community

Cover image for Introducing the Clocked Intelligence Hub - AI Integrations, Document Processing, and Workflow Automation in One Place
Ranjan Dailata
Ranjan Dailata

Posted on

Introducing the Clocked Intelligence Hub - AI Integrations, Document Processing, and Workflow Automation in One Place

Source Code

Download Source Code


📋 Prerequisites


Introducing the Clocked Intelligence Hub

A unified ecosystem for Clocked AI integrations, intelligent document processing, and workflow automation.

Building AI-powered business processes usually means stitching together separate repositories: one for your AI client integration, another for your automation workflows, and yet another for your document-processing application. The Clocked Intelligence Hub brings those worlds together in one place - without forcing every project into a single release train.

Whether you're an AI engineer connecting assistants to Clocked APIs, an automation engineer orchestrating n8n workflows, or a finance team processing invoices at scale, the Hub gives you three complementary integration surfaces built around the same Clocked AI capabilities.

🔗 Repository: https://github.com/ranjancse26/clocked-platform

📄 API docs: https://clocked.rezolve.ai/docs


Key Features

1. AI Integration Through MCP

The Clocked MCP Server provides an integration layer between compatible AI clients and supported Clocked API capabilities.

Depending on the implemented tools and configuration, this integration can support:

  • Structured information extraction.
  • AI-assisted decision-making.
  • Document processing operations.
  • Security and validation workflows.
  • Reusable AI capabilities exposed through MCP tools.

The server acts as an integration layer rather than replacing the underlying Clocked APIs.

2. Workflow Automation with n8n

The n8n-nodes-clocked project provides custom n8n nodes for invoking Clocked APIs directly from visual workflows.

The integration is designed to make AI capabilities available within existing automation pipelines without requiring every workflow to implement its own HTTP integration logic.

Potential benefits include:

  • Reusable workflow operations.
  • Centralized API authentication configuration.
  • Structured request and response handling.
  • Easier integration with webhooks and external systems.
  • Support for multi-step business processes.
  • Simplified workflow development and maintenance.

3. Intelligent Invoice Processing

The clocked-invoice-intelligence project applies AI-assisted processing to invoice documents.

The intended processing pipeline combines document parsing, structured extraction, security checks, financial validation, and decision-making.

Depending on the implemented workflow, the system can help:

  • Extract invoice fields into structured JSON.
  • Process supported PDF, image, or structured invoice formats.
  • Identify potentially malicious document instructions.
  • Validate extracted values against business rules.
  • Reconcile line items, subtotals, taxes, and invoice totals.
  • Identify discrepancies requiring human review.
  • Return structured processing results to downstream systems.

Financial validation should use deterministic calculations and explicit business rules wherever possible. AI-generated interpretations should not replace accounting controls.


What's Inside the Hub

The repository organizes three projects with distinct responsibilities, each independently versioned and deployable:

Project Purpose Primary Users
Clocked MCP Server Exposes Clocked capabilities through the Model Context Protocol AI assistants, agents, MCP-compatible clients
n8n Nodes for Clocked Integrates Clocked API operations into n8n workflows Automation engineers, workflow developers
Clocked Invoice Intelligence AI-powered invoice processing application Finance teams, developers, enterprise automation

1. Clocked MCP Server - AI Integration Made Simple

The clocked-mcp-server/ project is a stateless TypeScript MCP server that wraps the Clocked Fill, Extract, and Decide APIs over Streamable HTTP. Instead of calling Clocked REST endpoints directly, MCP-compatible clients (n8n, Claude Desktop, Cursor, and other MCP hosts) discover and invoke three standardized tools:

MCP Tool Clocked Endpoint Purpose
clocked_fill POST /clocked/fill Extract, classify, score, or generate structured fields from text
clocked_extract POST /clocked/extract Pull specific entities/values from form-like text
clocked_decide POST /clocked/decide Answer decision questions (approvals, priority, affected systems)

Key design principles:

  • Stateless - each request carries its own context
  • Auth separation - MCP_ACCESS_TOKEN protects the endpoint; CLOCKED_API_KEY signs outbound Clocked calls
  • Idempotency support via Idempotency-Key headers

2. n8n Nodes for Clocked - Visual Workflow Automation

The n8n-nodes-clocked-full/ project adds custom Clocked nodes to the n8n editor, so you can invoke Clocked APIs inside visual automations without hand-writing HTTP requests for every operation. Benefits include reusable workflow operations, centralized API authentication, structured request/response handling, and easier webhook integration.

3. Clocked Invoice Intelligence - Intelligent Document Processing

The clocked-invoice-intelligence/ project is a Next.js + React application that turns incoming invoices into structured, validated data downstream financial systems can trust. The stack includes:

  • Next.js 15 + React 19 frontend
  • Unstructured.io Transform for document parsing (PDF, images, structured formats)
  • Clocked Injection guard to screen parsed content for prompt-injection attempts
  • Clocked Extract for structured invoice field extraction
  • Clocked Decide for categorization and outcome recommendations
  • Zod extraction schema + deterministic TypeScript validation (financial reconciliation, duplicate checks)
  • PostgreSQL + Prisma for invoice records and processing history
  • Clerk authentication
  • Vitest financial validation tests
  • Demo mode so the UI can be explored without API credentials

The pipeline: upload - parse - security check - extract - validate - decide - approve or route for review. The app deliberately stops at a recommendation - it never authorizes or initiates payment.


Architecture

The Hub separates integration interfaces from business applications, letting Clocked APIs be accessed through multiple surfaces:

Clocked Platform Architecture

Architectural principles:

  1. Separation of concerns - each project has a defined responsibility
  2. Independent deployment - applications and integrations release separately
  3. Reusable capabilities - Clocked APIs accessed through different integration surfaces
  4. Security by design - credentials, document inputs, and processing results require safeguards
  5. Deterministic validation - financial calculations are independently verifiable
  6. Human oversight - exceptions route for review when automation can't establish a reliable result

How the Projects Work Together

Three common scenarios show how the Hub fits into real work:

Scenario 1: AI assistant integration

An MCP-compatible AI client connects to the Clocked MCP Server, discovers clocked_fill/clocked_extract/clocked_decide, invokes a tool, and receives a structured response - without implementing each Clocked API integration itself.

Scenario 2: Automated business workflow

An n8n workflow triggered by a webhook prepares input, invokes a Clocked node, evaluates the response, and routes results to another node, external system, or review queue.

Scenario 3: Invoice processing

An invoice enters the Invoice Intelligence app:

  1. The application receives the invoice
  2. A document parser extracts text and structural information
  3. Security checks inspect content for injection attempts
  4. Clocked Extract produces structured invoice fields
  5. The app validates required fields and financial calculations
  6. Clocked Decide evaluates the processing outcome
  7. The invoice is approved or routed for human review
  8. The app returns a structured result and records processing information

Demo

See the Clocked Intelligence Hub in action across all three projects.

Clocked MCP Server

The MCP server playground demonstrates tool discovery and invocation against the Clocked API.

Clocked MCP Server Playground

Reference documentation for MCP tools, transports, and supported capabilities.

Clocked MCP Server Documentation

n8n Nodes for Clocked

The custom Clocked node appearing in the n8n editor, ready to be configured in a workflow.

n8n Clocked Custom Node

An end-to-end invoice processing workflow built using the custom n8n nodes.

Clocked Invoice Processing Engine n8n

Clocked Invoice Intelligence

The Invoice Intelligence web application interface for uploading and processing invoices.

Clocked Invoice WebApp

Structured invoice parsing and extraction results displayed in the application.

Clocked Invoice Parse

Processing history and audit trail for invoices handled by the application.

Clocked Invoice History WebApp

Invoice being processed through the extraction and validation pipeline.

Clocked Process Invoice

Invoice routed for human review when automated processing cannot establish a reliable result.

Clocked Process Invoice Human Review

Note: These screenshots illustrate the target user experience. Actual layouts, fields, and behavior should match the implementation in each project.


Getting Started

Prerequisites

  • Node.js 20+ (Node 22 LTS recommended)
  • npm
  • PostgreSQL 15+ (for Invoice Intelligence)
  • A valid Clocked API key

Clone and explore

git clone https://github.com/ranjancse26/clocked-platform.git
cd clocked-platform
Enter fullscreen mode Exit fullscreen mode

Clocked MCP Server

cd clocked-mcp-server
npm install
cp .env.example .env   # set CLOCKED_API_KEY
npm run dev
# MCP endpoint: http://localhost:3000/mcp
Enter fullscreen mode Exit fullscreen mode

n8n Nodes for Clocked

cd n8n-nodes-clocked-full
npm install
npm run build
# Install the built package into your n8n instance
Enter fullscreen mode Exit fullscreen mode

Clocked Invoice Intelligence

cd clocked-invoice-intelligence
npm install
cp .env.example .env   # set DATABASE_URL and provider keys
npx prisma generate
npx prisma migrate dev --name init
npm run dev
# Open http://localhost:3000
Enter fullscreen mode Exit fullscreen mode

Set DEMO_MODE="true" to explore the Invoice Intelligence UI without sending documents to external services (built-in fixture data is used for walkthroughs).


Why It Matters

The Clocked Intelligence Hub removes the fragmentation between AI integration, workflow automation, and document processing. You get:

  • One repository for all Clocked integration surfaces
  • Multiple entry points - MCP for AI clients, n8n for automation, a dedicated app for invoice processing
  • Independent lifecycles - each project releases on its own schedule
  • Production-conscious defaults - deterministic financial validation, human-review routing, security screening, and credential separation

Get Involved

The Clocked Intelligence Hub is open source under the MIT license. Contributions are welcome - review the relevant project README, create a feature branch, implement your change, add or update tests, and open a pull request.

🔗 Star and fork: https://github.com/ranjancse26/clocked-platform

📖 Clocked docs: https://clocked.rezolve.ai/docs

Top comments (0)