# Agent Flow Bind: ready-made AI agents for repetitive work > The platform I designed and built: AI agents that read documents, answer the phone and check the administration, from the first page to the last approval. Every run is traced. - URL: https://mojtabaamini.com/projects/agentflowbind-en - Language: en - Author: Mojtaba Amini, AI Software Engineer (Verona, Italy) - Details: Product · 2026 — present · Verona ## The live demo On the Agent Flow Bind website you try the agents on a real sector document, or on your own: 3 free runs a day, no sign-up. The live demo on agentflowbind.com: pick the agent, then the document, then watch the run. 1. **Pick the agent**: By sector: electrical contractors, law firms, finance and administration, quotes. 2. **Pick the document**: A sample document from the sector, or your own. 3. **Watch the run**: The agent reads, extracts and checks, and returns the result with every step traced. ## What it does, sector by sector - **Electrical contractors · 4 agents**: Delivery notes checked against the order line by line, bills of materials for the wholesaler, weekly job sheets with hours and mileage, end-of-works and subcontractor files. - **Law firms · 1 agent**: A register of contracts, policies and deadlines: tacit renewals, notice periods and the clause each date comes from, with an .ics calendar. - **Finance and administration · 12 agents**: Supplier-invoice checks, FatturaPA and XRechnung e-invoices, bank reconciliation, reminders, receipts, expense claims, monthly summary and cash flow. They read and recommend: they never move money. - **Quotes · 3 agents**: Quotes from the drawing with your price list, supplier offer comparison, electrical drawings. - **Phone · 3 agents**: A secretary that answers, takes messages and routes urgent calls; outbound confirmation calls; an assistant that books on your calendar. agentflowbind.com ## Real scenarios, step by step Every scenario starts from a client’s situation, shows the files handed to the agent and the four phases of the run: diagnose, assemble, act, assess. Names and VAT numbers are invented; the structure of the documents is not. A scenario on agentflowbind.com: a delivery note photographed on site, the files and the agent’s phases. - [The delivery note, photographed on site](https://agentflowbind.com/scenari/delivery-note/consegna-parziale) (Scenario ↗): Two lines short, one over and a package under reservation against order OA-2026-0311. - [The invoice with one IBAN too many](https://agentflowbind.com/scenari/supplier-invoice/iban-di-troppo) (Scenario ↗): The supplier announces a new IBAN in the text, but the payment details still carry the old one. - [The deferred invoice: thirty-six lines over three delivery notes](https://agentflowbind.com/scenari/e-invoice-fatturaPA/differita-tre-ddt) (Scenario ↗): 22% and 10% bands, reverse charge, line discounts and three instalments. - [Manual at revision A, drawing at revision B](https://agentflowbind.com/scenari/bom-builder/manuale-rev-a-e-tavola-rev-b) (Scenario ↗): Two revisions that disagree: what actually gets ordered from the wholesaler. - [Office closed for lunch, a call about an invoice](https://agentflowbind.com/scenari/reception/messaggio-su-una-fattura) (Scenario ↗): The phone agent takes the number, the contact and the reason for the accounts office. - [A dental practice books a first appointment](https://agentflowbind.com/scenari/booking/studio-dentistico-prima-visita) (Scenario ↗): At 8:30, before opening, the assistant works out the request, checks the diary and books the slot. ## How it is built - **Agent kernel** with a diagnose → assemble → act → assess loop, tools, full tracing, budgets and human approval for anything that leaves the company. - **Models from several providers** (Claude, Gemini, OpenAI) behind one gateway. - **Channels**: web console, public REST API with a TypeScript SDK, webhooks and published web apps for end customers; telephony on Twilio, SIP and ElevenLabs. - **Data**: runs in the European Union, plus a local runner that reads a folder on the client’s own machine, so documents never leave it. - **Quality**: an eval set for every agent as a CI gate, and recorded tests in production. - **Stack**: TypeScript monorepo (pnpm and Turborepo), Next.js, Postgres, job queue, pay-per-run credits with Stripe. ## What it shows, for your company Agent Flow Bind is concrete proof of the services I offer: the same building blocks adapt to your company’s documents, calls and processes. - [Automatic document data extraction](https://mojtabaamini.com/services/document-data-extraction) (Document AI): From PDFs, invoices, contracts and scans to verified data in your ERP, CRM or Excel. - [AI voice and phone agent](https://mojtabaamini.com/services/ai-voice-agent) (Voice AI): Answers calls, qualifies leads, books appointments and updates your CRM. - [AI business process automation](https://mojtabaamini.com/services/process-automation) (Automation): Data entry, approvals, checks and reports automated inside your ERP and CRM. ## Frequently asked questions **Who built Agent Flow Bind?** I designed and built it: Mojtaba Amini, AI Software Engineer in Verona, Italy — product, architecture and code. **Can I try it for free?** Yes: the website has a live demo with 3 free runs a day, no sign-up, on sample documents or your own. **Where are documents processed?** In the European Union. Alternatively, the local runner processes a folder on the client’s own machine and the documents never leave it: only counts are reported back. **Can I get a custom agent?** Yes: templates can be copied and edited in the console, and I can build custom agents for your company’s processes. ## Let’s talk about your project Tell me about the process you want to improve: I will reply with a first assessment of where AI can help and where to start.