DE
Answering knowledge questions
Precise answers from your sources — with a citation.
See an exampleAgents and workflows
Hand over tasks,
keep control.
Agents take on recurring work across your knowledge. They stay visible step by step and put sources into the result — with the permissions you already have.
Your permissions, your rules
Workflows you can watch
Results with their sources
Tasks agents take on
The work that comes round every week.
Whether contracts, policies or mail — your agents take on the recurring work reliably, with a source, and inside the permissions you already have.
Putting research together
Structure information from several documents.
See an example
Recurring checks
Workflows on a schedule — completeness and inventory checks included.
See an example
Preparing handovers
Pull together where a case stands, so cover can carry on.
See an example
Keeping watch on collections
Spot changes and report them in time.
See an example
Building your own workflows
Set up recurring workflows from vetted building blocks.
See an exampleHow your agents work
Four steps
to the result.
From the idea to a running agent — configured simply, visible in the workflow, fully checkable.
- 1
Set the goal and the sources
You describe the task and choose the relevant sources (for example contracts, policies, mailboxes).
- 2
Configure the agent
You set the rules, the checks and the output format — with no programming at all.
- 3
Test and refine
The agent works through a sample. You review the result and adjust it if you need to.
- 4
Go live
The agent takes on the task automatically — on a schedule, or when new documents arrive.
An agent is not a master key
Safe. Controllable.
Under your supervision.
Agents work with existing permissions, keep every step on record and leave sources visible. What applies to your people applies to every agent.
Existing permissions apply
An agent sees only what the person who asked for it may see. There is no account of its own that reaches further.
Every step can be followed
Actions, decisions and linked sources stay visible — even months later.
Clear limits
You decide what an agent may do and what it may not. Reading and reporting change nothing in your systems.
Controllable at any time
Runs can be inspected, adjusted, paused or stopped. Anything that writes waits for your approval.
Common questions
Can an agent like this
break something?
Does the agent reach data without permission?
No. Every run takes on the existing permissions of the person who asked for it. Anything not released to you appears in no step and no result.
Can the agent return a wrong result?
A result carries its sources. You check it against the original, not on trust. If something is off, you see the passage and can stop the run.
How are changes recorded?
The plan, the steps, the sources and the result of every run are logged and can still be followed months later.
Who can see what the agent did?
Anyone who may see the run and the log under your permissions. The plan is open before the work starts, and every step carries its sources.
Can I stop the agent at any time?
Yes. Runs can be inspected, adjusted, paused or stopped. Actions that reach into a system wait for your personal approval anyway.
Which tools can an agent use?
Agents reach tools through vetted MCP servers. Tools that read work without asking; tools that act need your approval first.
Agents and workflows
Which task will you hand over first?
Think of a task that comes up every week in your team. In the trial you see how an agent plans it, which sources it brings back and at which point it needs your approval.