Coffee machines
Self-service troubleshooting for end customers, with QR codes on each unit and remote diagnostics.
- Reduces field interventions by 60%
- Step-by-step descaling guides
- Spare parts identification
Connect knowledge.Act intelligently.
More than vector search over your PDFs. Hybrid retrieval, custom chunking and compiled knowledge that learns from every interaction.
7-day free trial. No credit card needed.
- Hosted in the EU
helpcode.ai · troubleshooting
Hi! I'm your virtual support agent. What's the issue?
My espresso machine doesn't heat up.
Got it. Based on the manual and similar cases, let's check the heating element. Sources:
manual.pdf · p.42 FAQ · #18
Type your message…
Trusted by
Platform
Four products. One knowledge base.
Each product plugs into the same knowledge base, multi-agent orchestrator and analytics layer. Mix and match what you need.
AI troubleshooting
Guided decision trees that adapt to each customer in real time, generated automatically from your manuals and historical tickets.
- QR code or serial entry on every device
- Step-by-step diagnosis with photo and voice input
- Auto-escalation to a ticket with full context
Knowledge base
A dynamic knowledge layer with semantic search, citations linked to the original PDF and continuous retraining from real conversations.
- Connect SharePoint, Drive, Confluence, Notion
- Reranker + citations linked to the source page
- Gap detection and approval workflow
Learns from every fix
Your troubleshooting gets smarter, automatically
When one customer reaches a working solution, that path is reinforced for similar errors and similar devices — without manual labelling.
Solve once, scale to many. Every successful resolution becomes training signal: the next user with the same problem skips straight to the answer, and devices in the same family inherit the fix. Your support gets better the more it is used — a competitive moat most platforms don't have.
1 First customer solves at step 5 step 5
The AI walks through five branches before reaching the working fix on an espresso machine error E04.
2 Similar error, same model → step 3 step 3
The successful path is reinforced. The next user reaches the fix two steps earlier on the same family of devices.
3 Same error six weeks later → step 1 step 1
The fix is now the first suggestion. No manual labelling, no retraining ticket — the tree learned from real outcomes.
Knowledge that gets it right
An AI that knows your products like your best technician
We don't bolt a chatbot onto a folder of PDFs. We build a knowledge layer that learns your products, your jargon and your edge cases — so your customer gets the right answer the first time.
PDFs, schematics, tickets and tribal knowledge become a single source — and every successful resolution makes the next one easier.
The right answer, every time
Lexical, semantic and a reranker work together — your customer never lands on the wrong document.
Reads documents the way they were made
PDFs, schematics, tables, Confluence threads — every format ingested correctly, not flattened.
Compounds every week of use
From real conversations, the AI compiles concept pages your team can review and approve.
Speaks your specific domain
Error codes, part numbers, internal jargon — tuned per knowledge base, not generic.
Lexical (BM25)
+Semantic (dense)
→Merged
→Reranked
Hybrid retrieval blends lexical and semantic scores, then a reranker promotes the right page.
Model-agnostic
No vendor lock-in
Choose the best model for each task. Switch any time. Bring your own keys, run on-prem when sovereignty matters.
| Provider | Modality | Context window | EU region | On-prem |
|---|---|---|---|---|
 Claude Anthropic | Text + vision | 200K | EU | — |
 GPT OpenAI | Text + vision | 128K | EU | — |
 Gemini Google | Text + vision | 1M | EU | — |
 Mistral Mistral AI | Text | 128K | EU | |
 Llama Meta (open weights) | Text + vision | 128K | EU | |
 Azure OpenAI Microsoft | Text + vision | 128K | EU | — |
All models run in EU regions (Frankfurt via Azure, AWS Bedrock or GCP). Open-weight models run on-prem when sovereignty matters.
Switch model per channel, mix providers, run on-prem when sovereignty matters.
Built API-first
Plug helpcode.ai into anything you run
Every dashboard feature is also a REST endpoint. Issue an API key in two clicks and integrate with your CRM, helpdesk, mobile app or backend.
Versioned REST API
Stable /api/v1 endpoints for tickets, knowledge bases, vector search, items, organization — JSON in, JSON out.
API keys you control
Generate, scope and revoke keys from Dashboard → Settings → API Keys. Hashed at rest, shown once on creation.
Signed webhooks
Subscribe to ticket.created, kb.document.processed, session.handoff.requested. Each delivery is HMAC-signed for verification.
Built for production
EU-hosted, per-organisation isolation, request IDs and predictable error shapes. Generous rate limits, dedicated quotas on enterprise.
curl · helpcode.ai/api/v1
# Open a ticket from your backend
curl https://helpcode.ai/api/v1/tickets \
-H "x-api-key: hc_xxxx_xxxx" \
-d '{"subject":"Pump not priming","priority":"high"}'
# → 200 OK { "id": "9d2…", "status": "open" }Source Available · BSL-1.1
We built AnythingMCP because the gap was killing AI projects.
Every AI infrastructure stack stops at the model. Every knowledge base stops at retrieval. Between them sit your tools, APIs, ERP, ticketing — and that's exactly where most AI projects fail.
AnythingMCP turns any REST API into a native MCP server in minutes, no code. We open-sourced it because closing this gap should be table stakes for the industry — not a vendor moat.
AI infrastructure
Models & agents
Claude · GPT · Gemini · IDEs
The gap, closed
AnythingMCP
Any REST API → native MCP server, in minutes.
Your stack
Knowledge & systems
ERP · CRM · ticketing · KB
name: erp-bridge
upstream: https://erp.example.com/api/v1
auth:
type: bearer
token_env: ERP_API_KEY
tools:
- name: get_order
method: GET
path: /orders/{id}
- name: list_tickets
method: GET
path: /tickets?status={status}Drop a YAML file like this and the API is available to Claude, Cursor, helpcode and any MCP client.
- Bridges AI infra, knowledge bases and your systems
- MIT-licensed, self-hostable
- Native MCP — Claude, ChatGPT and IDEs out of the box
Enterprise-grade security
Your data, your rules
From day one helpcode.ai is built for enterprise compliance — EU-only data residency, GDPR by design and the deployment model that fits your stack.
GDPR
GDPR-compliant by design
Per-organisation isolation, encryption at rest, data export and erasure flows out of the box. Signed DPA available.
EU only
Hosted in the EU
Servers operated in Frankfurt, Germany. Customer data never leaves the European Union.
ManagedPrivate cloudOn-premises
Deploy on any infrastructure
Managed in our cloud, in your private cloud or fully on-premises — you choose where the data lives.
Use cases
Built for service teams
From facility management to electronics, helpcode.ai cuts ticket volume and frees technicians to focus on what really needs them.
Explore all use casesHVAC and heat pumps
Technicians on site get instant access to the right manual, the right schematic and the right past case.
- Voice-first interface for hands-busy work
- Schematic search across PDF manuals
- Compliance-ready audit trail
Ventilation systems
Filters, frost protection, controller fault codes — guide the building manager through every routine task and most fault codes.
- Filter change schedule by AHU model
- Heat-exchanger frost diagnostics
- Commissioning checklists with sign-off
Sensor technology
Pressure, level, temperature, flow, gas — diagnosis, calibration and documentation for the brand and model in front of the technician.
- Calibration walkthroughs by sensor type
- Wiring and EMI diagnostics with diagrams
- Calibration certificates auto-generated
FAQ
Frequently asked questions
Everything you need to know before you start. Can't find your answer? Drop us a line.
What is helpcode.ai and what makes it different?
An AI platform that connects your company knowledge. Troubleshooting, KB and Hotline share the same brain. The differentiator is the knowledge layer underneath — hybrid retrieval, custom chunking, domain-tuned embeddings.
What happens when the knowledge base has no answer?
The AI says so, openly. The session is flagged as a gap, escalated to a ticket with the full transcript, and surfaces in the dashboard's Gap detection view so your team can fill it once and answer it forever.
Which languages does helpcode.ai speak?
The dashboard and marketing site are in English, German, Italian, French and Spanish. The AI itself replies in the customer's language regardless of UI locale — including languages we don't ship a UI for.
What if the AI is wrong on a safety-critical case?
Every assistant ships with a domain-tuned safety prompt, automatic escalation triggers for keywords and confidence drops, and the option to gate any answer behind a human-in-the-loop. Citations to source pages mean disagreements are auditable, not opaque.
How does helpcode.ai integrate with my existing systems?
Native connectors for SharePoint, Drive, Confluence and Notion; a versioned REST API with signed webhooks; and AnythingMCP to wrap any REST API as an MCP server in minutes.
Why did you build AnythingMCP and is it really open source?
Yes — Source Available under BSL-1.1, self-hostable on GitHub. We built it because between models and your ERP/CRM/ticketing sits the gap where most AI projects fail — that should be table stakes, not a vendor moat.
What does it cost?
Trial is free. Business starts at €40/user/month (yearly) or €50 (monthly). Enterprise is custom — for 1,000+ users, dedicated deployment, or white-label.
Can I white-label helpcode.ai?
Yes. The chat widget, the dashboard and the email templates can all be re-branded with your name, your colours and your domain. Reseller margins are negotiated case by case.
More questions? Email us at info@helpcode.ai