Behind the scenes

The AI tools
the pros use

Here is our complete stack: which tools run at GO10X every day, how they fit together, and which tricks have proven useful. Take whatever fits your company.

  1. BuildClaude Code · Codex · Cursor
  2. ModelsClaude · GPT · Gemini · Mistral
  3. AgentsOpenClaw
  4. Data and knowledgeHubSpot · GitHub · PostgreSQL
  5. Automationcustom code

Which AI tools does GO10X use?

GO10X builds with Claude Code and picks the AI model for each task: Claude, GPT, Gemini or Mistral. Our CRM is HubSpot, and our knowledge and data live in Obsidian, GitHub and PostgreSQL. Anything that contracts or invoices depend on, we write as custom code.

How GO10X is set up

Where the data lives

  • HubSpotcustomers and orders
  • Microsoft 365email and files
  • E-signature and accountingcontracts and invoices
  • PostgreSQLour own database

Where the agents work

  • Claudebuilds and writes
  • ChatGPTclicks in the browser
GitHubKnowledge, rules and code as text files
  • Company knowledge
  • Automations
  • Website
Our own serverThis is where the automations run.
  • HubSpot automation
  • Content Engine
  • Apps

What customers see

Websitebuilt from text files
Inquiry formlands in HubSpot
A few systems hold the data. Automations connect them through APIs, and everything else lives as text files in Git.That is why we can change any part of it with AI.
From our day-to-day

Nine tips we give everyone

  • Work in a repository, not in a chat

    We work with Claude Code in Git repositories on GitHub. The files and the version history belong to us, and ChatGPT, Codex or Cursor can open the same repository.

  • Connect your accounts first

    A connector gives the chat access to email, calendar, CRM and files, with your permissions. Then the agent does more than summarize: it files, drafts the reply and logs everything in the CRM.

  • Set a goal, not steps

    Describe what the end result should be, and check it. The agent plans the steps itself.

  • One model plans, many agents work

    In the main chat, the strongest model plans, delegates and reviews. Subagents do the work, each starting with a clean slate, and for routine work like researching 20 companies, the cheapest model is enough.

  • Build lots of small skills

    A skill is a recurring task with a fixed input and a fixed result. One of our skills turns call notes into a finished profile as a PDF.

  • What should last belongs in files

    Whatever lives only in the chat history is gone in the next session. Our eleven contract templates live as text files in Git, get filled from HubSpot and are signed as PDFs.

  • Let the agent click in the browser

    For forms and portals without an API, ChatGPT uses the browser. An agent filled in our trademark application at the German Patent and Trade Mark Office.

  • Talk instead of typing

    We dictate tasks and ideas instead of typing them. An agent checks the dictation against email, calendar and CRM and reorders the priorities.

  • Track costs from day one

    Our code logs every AI call with its tokens and cost, and an email shows the spend every day. One day of our Content Engine, newsletter and podcast included, costs about 35 US cents.

For your company

Tool highlights

What we work with every day, and what we have built with it.

  1. ›Take the next item from TODO.md.
  2. Reads CLAUDE.md and the rules
  3. Two subagents review copy and code
  4. Type check passed
  5. Build passed
  6. SEO check passed
  7. Pull request opened, waiting for approval
Our most important tool

Claude Code

We use Claude Code every day and build internal tools, this website and new products with it, for example our License Check for HubSpot.

  • Team rules in CLAUDE.md
  • Skills and connections to your systems
  • Permissions set centrally
AI applicationClaude, ChatGPT or Copilot
  • HubSpotContacts, deals, notesRead, drafts for approval
  • Microsoft 365Mailbox, calendar, SharePointRead only
  • ERPCustom MCP serverWrite after approval
An open standard

MCP server

An MCP server connects AI to your systems, such as HubSpot or Microsoft 365. We decide what the AI may do in them and put the connection into production.

  • Official servers first
  • Permissions as tight as possible
  • Writing only with approval

What you use

  • Workflows: 24 of 38 activeused
  • Seats with a login: 8 of 11used
  • Sequences: 0 set upunused
Features per hubusedunused

Read-only, nothing changes in your portal

Built by us

License Check for HubSpot

The check reads your HubSpot portal once and shows which features of your plan your team uses and which sit idle. Then you know whether you can save money or get more out of your license.

  • Free
  • Read-only access
  • Report in one to two minutes
The stack in detail

Which AI tool for what?

We choose the tool task by task. Our knowledge lives in files and code that belong to us. So when we switch a tool, we do not start from zero.

Build

Claude CodeOur most important tool

We use it to write code, automations and this website. Its own page explains how we roll it out for your team.

Claude Code in your company

Codex and CursorWe use them too

Which of these tools fits depends on where your code lives, which development environment you work in, and what an agent is allowed to touch.

Codex for your company

Models

Claude, GPT, Gemini, MistralPicked per task

The strongest model makes the decisions, and a cheaper one researches and pre-sorts. Through OpenRouter we reach every model with a single key.

ChatGPT BusinessFor every department

ChatGPT's company plan: one workspace with roles, single sign-on and no training on your data. How we set it up and roll it out one department at a time is on its own page.

ChatGPT Business for your company

Agents

OpenClawAgents in the team chat

A platform for AI agents that several people work with in a chat. We built our lead research on it. We have paused the platform for now and are rebuilding the research as a skill.

Data and knowledge

HubSpotCRM

We automate inside HubSpot instead of building a second CRM next to it. Your team keeps working in the interface it knows.

Sales automation in HubSpot

Obsidian and GitHubKnowledge as text files

Priorities, decisions and rules live as text files in Git. An agent reads and writes them directly, with no API in between.

PostgreSQL and SupabaseDatabase

We keep company data in a database an agent can query directly. Our Content Engine runs on it.

Automation

Custom codeFor everything critical

When contracts, invoices or customers depend on a workflow, we write code. It is versioned and tested and sends a report every day.

Our view

What we advise against.

We see these four mistakes again and again in first calls.

  • Building the processes that contracts or invoices depend on as point-and-click workflows. Without tests and versioning, nobody notices when something breaks.
  • Buying licenses without rolling them out. The bill arrives every month, and after a few weeks nobody uses the tool anymore.
  • Giving every department its own tool. One department never learns what runs in the next.
  • Committing to one agent platform. The platforms overtake each other every month, while your files and your code stay.
FAQ

Common questions.

Claude or ChatGPT: which is better for a business?

Claude and ChatGPT are about even. We run both, plus Gemini and Mistral. The setup matters far more than the model: if your knowledge lives in files and code that belong to you, you can switch models without starting from zero.

Which AI tool should a small business start with?

Start with one workflow that keeps getting stuck in your day-to-day, and only then pick the tool. At the start we put three use cases in front of you, and one of them goes live.

Does our data stay in the EU?

Yes. Our servers are in the EU, and for the AI models there are three options: Azure OpenAI with a data center in Sweden, Mistral from France, or a local model that runs on your own machines.

How much do AI tools cost a business?

Vendors set their license prices, and those change often. That is why we do not list any here. The most expensive license is the one nobody uses. For HubSpot, our free License Check shows which features of your plan sit idle.

Free AI Strategy Call · 30 minutes

Which tool fits your bottleneck?

We look at your bottleneck and tell you which tool fits your company and what we advise against. The call is free.