Custom AI agent development
We build your agent.
Turn a repeated task into an AI agent for your team. Built around your knowledge, rules and tools. We set it up and support it after.
We reply within one working day. Happy to sign an NDA before you share anything.
- Knowledge
- Rules
- Skills
- Gmail
- ERPNext
- Slack
- Chat
- Workflows
- Your assistant · MCP
Pick the tools you use.
One shared brain. Reused across your agents.On open-source Hubzoid
Trusted by teams at
What we build
Start with a job your team knows.
We adapt a ready-to-use agent to your process.
Or
Have a different process? We build for it.
Business questions, document checks or recurring reports. Start with one.
/ask-the-businessChat · When someone asksAnswers from your business data
Plain-language answers on stock, sales or cash, with the source.
What did the main branch sell yesterday?Up on last week, led by weekend sales.Source: point of sale/document-checkChat · When a document arrivesDocument checks
Reads an invoice or slip and marks it passed, failed or needs review.
Supplier slipPassedFailedNeeds review/scheduled-reportWorkflow · On a scheduleScheduled reports
Pulls from your systems and sends each team its own summary.
LeadershipSection headsFloor huddle
How we work
One process. Built, tested and supported.
Agree the work first. Test it with your team. Support it after setup.
Your process. A working agent.
About 2 minutes
Watch on YouTube (opens in a new tab)Choose the task
Walk us through the process. Agree the scope, a useful result and the quote.
Build and test
We connect your knowledge and tools, then test on your examples with your team.
Set up and train
We deploy the agent and train the people who use and maintain it.
Support after setup
Maintenance, fixes and refresher training within an agreed support scope.
Built on open source
Your agent. Your infrastructure.
Built on Hubzoid, open source under Apache 2.0. Keep your Hub and the option to run it yourself.
- Editable knowledge, instructions and tools
- Deployment setup and operating documentation
- Your choice of supported model provider
- An agreed plan for ongoing support
Your infrastructureYour server, cloud account or laptop
Your HubYours
AGENTS.mdknowledge/skills/tools_local/
Your agent’s files, versioned in your Git repository.
Runs on
HubzoidApache 2.0
- Chat
- Workflows
- MCP
- Settings
Open source. Free to use, change and run.
Stored here: web chat history, runs and the access log.
Outbound only to what you configure
Your model provider and the systems you connect
Systems we connect to
Keep the tools you already use.
We scope the connections your process needs, including API, database and file access.
- ERPNext
- SAP
- Notion
- Zoho
- HubSpot
- Google Drive
- Gmail
- Outlook
- Slack
- Telegram
- YouTube
Access you control
Your team stays in control.
You control who can use each agent and which tools it can call.
- Manage access in Settings
- Restricted tools need an explicit grant
- Agree which actions need your team’s review
By industry
Find a starting point for your industry.
Explore the work an agent could help your team with.
Start with one task
Get a demo.
Tell us what your team does today and what you want to improve. We will reply by email to arrange a demo.
A founder reads every message.
- A setup fee, then support. We quote both after a short call
- We reply within one working day. Happy to sign an NDA before you share anything
Questions
What teams ask before we start.
More about the build and what happens after.
What kind of agents do you build?
We start from a ready-to-use agent, such as Order & Invoice Desk or Customer Desk, and adapt it to your rules. Or we build a new one around your process, such as answers from your data, document checks or scheduled reports.
What does it cost?
A setup fee to build and set up the agent, then support. We quote both after a short call about your process, systems and scope. Model and hosting costs depend on the setup.
We are not sure which agent we need. Where do we start?
Bring one repeated task or a process you want to improve. We can discuss where an agent could help and what it would need. You do not need a technical specification.
What happens after setup?
Your team uses the agent day to day. We agree the maintenance, fixes and refresher training included in support. Further changes are scoped together.
Who owns the agent and the data?
You keep your Hub, knowledge and agent instructions. Hubzoid is open source under the Apache 2.0 license, with a self-hosted route. Connected systems and model providers have their own terms.
Where do the agents run, and which AI models do they use?
In your cloud or on your own servers, with a model provider Hubzoid supports. That provider’s data terms apply.
What do you need from our team?
Someone who knows the process, representative examples and time to review results. We agree the system access needed for the build.
Which systems can you connect?
Most systems with an API, a database or a regular export. We check yours before we agree the scope.
We build your shared brain
Hubzoid