Products / Agent Core+
Agents do the work,
you keep
the approval
With Agent Core+ you build your organization's agents and workflows visually, connect them to your documents and run them on your own servers. It installs with one click and comes with support.
Problem
You want automation, not a loss of control
Handing repetitive work to AI is tempting. But an agent that writes to company systems on its own is unacceptable, and cloud agent platforms take your data outside. Agents that each team builds with its own tool cannot be monitored or governed.
What it does
- 01
Visual workflow builder
Build agents and workflows step by step on a drag-and-drop canvas.
- 02
Sourced knowledge search
Add your documents to a knowledge base; agents answer with document and page citations.
- 03
Tools and connections
Agents connect to your systems, services and MCP tools.
- 04
Approval steps
Steps that write to a system, send an email or change data come for approval first.
- 05
Triggers
Scheduled ("every morning"), event-based ("when an invoice arrives") or run by hand.
- 06
Ready-made templates
Start fast from templates for common agents.
- 07
Monitoring and logs
Every run's steps, durations and results are logged; errors on one screen.
- 08
Publish as API
The agents and workflows you build are exposed as APIs to your other applications.
How it works
01
Build
Design the agent or workflow on the canvas; connect knowledge sources and tools.
02
Set limits
Define permissions, approval steps and triggers.
03
Run
The agent runs on a trigger or a call; at a step that needs approval, it stops and asks.
04
Monitor
Every step is logged; results and errors on one screen.
Example agents
- 01
Invoice agent
Reads an incoming invoice, checks it against your rules, prepares the pre-posting and presents it for approval.
- 02
Policy assistant
Answers employee questions from company policies, with page citations.
- 03
Weekly report agent
Every Monday, gathers the data, prepares the report and sends it to the right people.
- 04
Request routing agent
Classifies incoming requests and forwards them to the right team.
Product screens
01 · Agent builder and run view with an approval step
02 · Knowledge base and source preview
03 · Agents and monitoring
04 · New agent from a template
Principles
- 01
Approval first
Every step that writes to a system comes as a card first.
- 02
Under its own permissions
An agent can never do more than the person or role running it is allowed to.
- 03
Every step logged
Which agent did what, when, and with whose approval: all on record.
- 04
On-premise
Agents reach models from inside your organization, through Bizz AI Gateway.
Installation and support
Installs on your servers with one click; no customization needed.
We provide installation, initial training and support.
Available as part of the Bizz AI Core package or on its own.
FAQ
Who builds the agents?
Your authorized users, with the visual builder; no coding needed.
Can an agent do something without approval?
No. Any action with an approval step is never sent until it is approved.
Which models does it work with?
Any model connected to Bizz AI Gateway; on-premise models are the default.
Can it connect to our own systems?
Yes; through services, APIs and MCP tools.
How is it installed?
On your servers with one click; installation and support come from us.
Agent Core+
Want to see it on your own servers?
We will show our products against your needs, using fictional data.