[HIRING] Claude API + n8n Engineer — 31-Agent Business Intelligence System | Remote

Jobs
1

Looking for an experienced Claude API + n8n engineer for a serious production build.

THE PROJECT
31-agent intelligence system for a multi-entity portfolio: hotels, jewelry, wine and spirits retail. Caribbean-based. Mac Studio M4 Max 128GB dedicated machine. All agents via Claude API key.

AGENT EXAMPLES

  • Weekly sales brief pulling from QuickBooks
  • Hotel rate monitor via Duetto API
  • Source verification filter between search and strategic analysis layers
  • Multi-source intelligence synthesis at L3

STACK

  • n8n self-hosted on macOS
  • Python scheduled via cron
  • Claude API (Sonnet for L1/L2, Opus for L3 with extended thinking)
  • QuickBooks, Duetto, Sage Intacct, Gmail

WHAT MATTERS

  • Production Claude API experience — not demos
  • n8n self-hosted workflows shipped to production
  • QuickBooks integration experience
  • Clean credential management
  • Async, documented, operator-friendly handoff
  • Comfortable working with wine and spirits data

ENGAGEMENT

  • Phase 1: 4 agents, $
8 Likes
2

This is a very interesting build. I have 3+ years of experience working with n8n and AI-driven automation systems, including building production-level workflows, API integrations, and multi-step logic pipelines. I’ve worked with LLM-based systems using OpenAI/Claude for structured outputs, data processing, and business automation use cases.

I’m also experienced with self-hosted n8n setups, clean credential handling, and building scalable, well-documented workflows for real-world operations.
I’m comfortable working on complex, multi-agent architectures like this.

Check Inbox Happy to connect and discuss further.
Email: muhammadmoosa.abc1@gmail.com

1 Like
3

Hey :waving_hand:,

I’m Milan, with 8 years of experience in Business Automation and AI. Including 2 years at Apify working on enterprise-level browser automation.

Currently specializing in n8n, but also proficient in Python & Javascript.

Find out more about my work here:

If you think I might be a match, please:

Book a Call with me

Or reach out at hello@smoothwork.ai

Looking forward to hearing from you!

2 Likes
4

Hi, i am interested in this project i have 3 years of experience in AI and automation from Infosys India. Has a strong educational background which includes tier 1 indian engineering college, I have 1 year of experience with N8N for automation, this project seems interesting it will be great if i will get chance to work on this project.
Thanks and Regards
AMIT
Ph no. 9602947279
email amiteng5472@gmail.com

5

Hey Profile - Trefle - n8n Community,

I got you, I have been building all forms of automations for the past 2 years and have built 100s of flows for my clients. Have worked with all sorts of companies and gotten them 10s of thousands in revenue or savings by strategic flows. When you decide to work with me, not only will I build this flow out, but also give you a free consultation like I have for all my clients that led to these revenue jumps.

I have built a similar workflow like this for one of my clients. I can not only share that but also how you can streamline processes in your company for faster operations. All this with no strings attached on our first call.

Here, have a look at my website and you can book a call with me there!

Talk soon!

6

Hi Trefle,

This one caught my attention — I recently built and deployed a 9-agent intelligence system for a UK-based e-commerce business running 6,000+ products. Same architecture pattern you’re describing: Claude API (Sonnet for routine analysis, Opus for complex reasoning), Python orchestration, and n8n handling scheduling and triggers.

A few specifics that map directly to your setup:

  • Each agent owns one domain (ad waste detection, pricing optimization, executive briefings) and writes structured observations to a shared context layer in Supabase. Agents cross-reference each other’s findings — so when the Ad Waste Detector flags a problem, the Pricing Optimizer picks it up automatically.
  • Claude Sonnet handles L1/L2 classification tasks. Opus handles the deep multi-step analysis — weekly CEO briefings with strategic recommendations delivered to Slack.
  • API orchestration across 7 data sources (Amazon Ads API, Keepa, ClickUp, and others), all running on automated schedules.
  • Identified $30K+ in wasted ad spend in the first month. Full case study: priyanshukumar.co/work

Scaling from 9 agents to your 31-agent system is an architecture challenge I’d genuinely enjoy. The shared context layer pattern I’ve built scales cleanly — the 10th agent reads the same observation table as the first.

Happy to walk you through the architecture in more detail — or share a Loom of the system running.

Priyanshu Kumar
AI & Automation Engineer
priyanshukumar.co

7

I’m very interested, I have over 5 years of experience as a Software Engineer, I have worked with aws,node,express,mongodb,react ,python ,docker and have done several projects using n8n and claude code mcp sdk api and hooks

8

Hi Priyanshu,
Thanks — the Supabase shared context layer is exactly the pattern we’re working with.
Three quick questions before we go further:

  1. Fixed price for Phase 1: 4 agents, tested, documented, deployed — what’s your number?
  2. The system integrates QuickBooks, which runs behind a VPN. How do you handle secure remote access to a client’s internal systems?
  3. Our strategic agents require Claude Opus with extended thinking enabled. Have you built with Opus specifically, and have you worked with extended thinking mode?
    If all three check out, we’ll schedule a 20-minute call this week.
    Best,

Trefle

9

Hey @Trefle ,

This is a serious build — and honestly, most people here will underestimate the complexity of a 31-agent system.

The challenge isn’t just “connecting APIs” — it’s structuring reliable agent layers (L1/L2/L3), controlling context flow, and preventing hallucinated or low-quality synthesis at higher levels.

That’s exactly the kind of systems I work on.

How I’d approach this:

  • Agent Architecture
    Design clear separation between:

    • L1 → data extraction (QuickBooks, Duetto, APIs)

    • L2 → structured processing + validation

    • L3 → synthesis (Claude Opus with controlled context + source verification)

  • n8n Orchestration
    Build modular workflows (not spaghetti) with:

    • queue handling for async execution

    • retry + failure isolation

    • clean credential management

  • Claude API (Production Use)
    Not just prompts — I implement:

    • structured outputs (JSON enforced)

    • context window control for multi-agent chaining

    • cost + latency optimization between Sonnet / Opus

  • Data Integrity Layer
    Especially important for financial + BI use cases:

    • source verification before synthesis

    • deterministic checks before L3 agents

    • logging + traceability per agent

Relevant work:

I’ve built multi-step AI systems using n8n + LLMs for:

  • automated reporting pipelines

  • lead intelligence + enrichment systems

  • AI-driven decision workflows across multiple data sources

Links:

:globe_with_meridians: https://www.muhammadz.fun/
:brain: https://www.notion.so/muhammad-ai-automations/AI-Solutions-Automation-Showcase-2026-2f8a292a24138082acece2ccbb1c3a3b

Contact:

:e_mail: muhammad.specials@gmail.com
:mobile_phone: WhatsApp: +92 3360327970

If helpful, I can map a clean 31-agent architecture draft before we even start Phase 1.

— Muhammad

10

Your multi-agent setup needs solid RAG infrastructure underneath it all. Without it, those 31 agents will hallucinate all over your business data.

For your QuickBooks/Duetto integrations, don’t push everything through Claude directly. Build a data pipeline that cleans and transforms first, then feeds structured data to the models.

Dealt with this exact pattern before: Multi-Agent RAG Platform | Foundry | Kingsley Onoh

Are you planning to implement any guardrails for the Opus-powered L3 agents? Those can get expensive fast without proper cost tracking.

11

Hi Trefle, Welcome to the community :waving_hand:

This sounds like a serious and well-structured system — especially the multi-agent architecture and the L1–L3 intelligence layering.

I’d be happy to collaborate on this.

I specialize in building AI agents and automation systems using n8n with API integrations and scheduled Python/automation workflows. Most of my work focuses on production-grade systems that handle data extraction, processing pipelines, and multi-step business logic.

I’ve worked on similar automation builds involving:

API-based data collection and processing

AI-assisted workflow orchestration

Multi-step business automation systems (CRM, email, and reporting flows)

From your description, I’m particularly interested in the combination of QuickBooks + Duetto + layered intelligence filtering — that’s a solid real-world use case for structured agent systems.

Here are a few relevant examples from my work:

• AI Voice Agent Automation (API + Vapi integration)

https://www.upwork.com/freelancers/~0122761e4734295f4b?p=2038586338272239616⁠�

• Multi-Channel Automation Workflow

https://www.upwork.com/freelancers/~0122761e4734295f4b?p=2039118619839795200⁠�

• Shopify Order Automation System

https://www.upwork.com/freelancers/~0122761e4734295f4b?p=2039336822638325760⁠�

I’d be open to a quick 10–15 min callto understand the architecture and see how I can contribute to Phase 1 of the build.

Best,

Folafoluwa Stephen

folafoluwaolaneye@gmail.com

12

Hi,

This is exactly the kind of production work I live for. 31-agent intelligence system across portfolio operations—I understand the complexity and why this needs serious, documented execution.

I’m a strong match:

  • Production Claude API shipped (Sonnet/Opus, extended thinking workflows)

  • n8n self-hosted pipelines in production (credentials, error handling, async ops)

  • QuickBooks + multi-source API integrations (Duetto, Sage Intacct, Gmail)

  • Python orchestration via scheduled tasks

  • Comfortable with structured data domains (finance, hospitality, retail)

What resonates: You’re looking for operator-friendly handoff and clean credential management. That’s exactly how I build—documentation first, so your team maintains it confidently.

On wine & spirits data: I’ve worked across retail inventory and supplier intelligence systems. The data patterns are familiar.

I’d love to discuss Phase 1 (4 agents) and your vision for the full 31-agent ecosystem.

Check out my work at ahmed-ali.dev — you’ll see production integrations, n8n workflows, and Claude API implementations.

Lets build together,

Ahmed Ali

13

Trefle,

Six weeks from now, your 31-agent system is live. Hotels, jewelry, wine, each feeding their own data stream into an Opus synthesis layer that drops a weekly intelligence brief in your inbox. You read it Monday. Nobody touched a spreadsheet.

I run this stack for a jewelry retail operation right now. Claude API through n8n, Sonnet on the lower layers, Opus on synthesis. Credential isolation across multiple business types is solved territory for me: separate stores per entity, shared orchestration above. That call has to happen at agent five, not twenty-five.

The L3 prompt chain is where I’d invest real design time. I’d need to see what L1/L2 actually outputs before touching it. Synthesis quality lives entirely on what structure arrives at that layer.

First build as a fixed milestone. You see it running before committing to the full scope. 30-day guarantee on everything delivered.

What does your current setup look like? Even rough.

Daemon

14

Ahmed,
Thanks for reaching out — your background is relevant and the stack alignment is real.
A few things before we go further:

  1. GitHub — please share your profile or 2–3 repos that best represent production Claude API and n8n work. The site is noted but I need to see code.
  2. Quick scoping question: describe how you would structure credential isolation across 4+ agents that each need access to different external APIs — without any agent having visibility into another’s keys.
    Keep it brief. I’m looking for your mental model, not a full design doc.
    If the answers are solid, I’ll share Phase 1 scope and we can discuss terms.
    Louis
2 Likes
15

Hi Muhammad,
Thanks for the detailed breakdown — the L1/L2/L3 framing and your points on context control and data integrity are on point for what we’re building.
Before we go further, a few things we need from you:

  1. GitHub profile link — we want to see actual deployed code, not showcase pages
  2. One live project example — specifically anything involving financial data pipelines, multi-agent orchestration, or structured output enforcement at scale
  3. Your rate — hourly or project-based, in USD
    A few things to know about this engagement:
    • The system is 33 agents across 5 tiers, not 31 — architecture is already defined
    • Data sources include hospitality PMS, accounting (Sage Intacct, QuickBooks), and retail POS
    • This is a USVI-based operation — vendor compliance matters
    • Phase 1 scope will be defined; we are not starting from scratch on architecture
    If the above fits, send us your GitHub and project example and we’ll take it from there.
1 Like
16

Daemon,
Good timing. We’re mid-build on a multi-entity stack — hospitality, retail, real estate — so the credential isolation problem is live for us right now.
Before a call, three quick questions:

  1. Fixed-price Phase 1: how do you scope and price it, and what’s the specific deliverable the client signs off on?
  2. Security posture: how do you handle VPN access and accounting system credentials (QuickBooks, Sage) across client environments?
  3. Opus extended thinking: describe a synthesis layer you’ve built where it materially changed output quality — what was the prompt structure?
    If those answers land well, I’m happy to get on a call and walk you through the current setup.
    Trefle
17

Hi Trefle,

Thanks for the quick follow-up. Happy to answer on VPN and Opus architecture here — pricing and scheduling I’ll send you via DM so we can get into specifics without spamming the thread.

VPN / QuickBooks access: I follow the client’s existing security posture. For QuickBooks specifically, the cleanest pattern is running the data ingestion agent inside your VPN boundary — either self-hosted n8n or a Python worker on your Mac Studio. Raw credentials and financial data never leave your network. Only structured, anonymized observations get written to the shared context layer for cross-agent consumption. If you use Tailscale, WireGuard, or a corporate VPN client, I operate through that. I don’t bypass or tunnel around client security.

Opus + extended thinking: Yes on Opus for strategic reasoning. I’ve tested extended thinking across different prompt structures. For your L3 agents synthesizing across hotels, jewelry, and wine & spirits, extended thinking is the right call — multi-entity portfolio analysis benefits from deep reasoning chains.

My approach:

  • L3 synthesis agents — Opus with extended thinking enabled. Calibrate thinking_budget empirically, start around 8-10K tokens, measure output quality vs latency.
  • L2 cross-entity analysis — Opus without extended thinking or Sonnet, depending on latency budget. Smart enough for multi-source reconciliation.
  • L1 classification/routing — Haiku or Sonnet. Extended thinking adds latency without improving pattern-matching — wrong tool for this layer.

The goal is matching compute tier to reasoning depth — not blanket-using Opus everywhere.

Sending pricing and scheduling via DM.

Priyanshu

18

Hi Trefle,

This is a strong match for what I do daily. I build production systems with Claude API (Sonnet for fast tasks, Opus with extended thinking for complex analysis) orchestrated through n8n and Python.

Relevant experience:

Multi-agent prediction system — Built and maintain a monorepo with 4 active bots across Manifold, Kalshi, and Polymarket. Each uses specialized data-driven evaluators (crypto via CoinGecko, weather via Open-Meteo, sports via odds APIs) dispatched through a classification layer. Runs on GitHub Actions with SQLite tracking and Telegram alerts. Architecture is similar to your L1/L2/L3 tiered approach.

n8n + API orchestration — Currently building automated business workflows for an e-commerce client: WhatsApp chatbot, order notifications, abandoned cart recovery, all through n8n connected to Tiendanube webhooks + WhatsApp Cloud API.

Claude API daily user — I build with Claude Code as my primary development tool. Comfortable with prompt engineering, token optimization, and choosing the right model tier for each task.

I can share my GitHub (private repos — happy to grant access) and walk through architecture decisions async via documented write-ups.

For Phase 1 (4 agents), I’d be comfortable with a fixed-price arrangement. What’s the budget range you’re targeting?

Best,
Diego Ostrovich

diegonoseencuentra@gmail.com

GitHub: Ostro1982 · GitHub
LinkedIn: https://www.linkedin.com/in/diegoostrovich/

19

Hi! this project is exactly what I work on.

I build multi-agent n8n systems with layered architectures, Claude API (Sonnet + Opus), structured outputs, and operator-ready documentation. My most recent project cut a client’s manual workload 65% in 45 days. Portfolio: gonzalo-insua.vercel.app

Happy to discuss scope and details over DM or email.

— Gonzalo | gonzalo.insua11@gmail.com | Buenos Aires (GMT-3)

20

Hi Priyanshu,
Your VPN and Opus answers are solid — exactly the right approach for this build.
Ready to move forward. Next step is a mutual NDA and service contract before we exchange specifics.
Please send me your standard NDA and service agreement, or let me know if you’d prefer to work from ours. My direct email is [your email] — let’s continue off-platform from here.
Best,
Louis