Security and trust
When you create an agent you choose the knowledge it may read, the tools it may use and the agents it may hand work to. The rest of this page follows from that.
What an agent may reach
A support agent sees the refund policy and the product notes. It does not see the board pack.
Where a person stands
Two separate mechanisms. One holds work an agent has produced. The other moves a live conversation to a colleague.
A draft reply, an invoice, a plan for the next agent. Work you mark for approval waits for a person, with the agent's reasoning beside it. Approve it, edit it, or send it back with a reason.
When a customer asks for a person, the agent moves the conversation to a shared queue and says so. A colleague picks it up on the same channel, and can hand it back.
The record
The documents the agent read and the tools it used, in order, with what came back. Open last month's run and read it.
See what changed, who wrote it, and put a known good version back.
Who approved what, when, and the reason attached to anything sent back.
Data handling
Conversations show what customers ask. Before a transcript is saved as knowledge, personal data is replaced with placeholders, so the shape of the conversation is kept without the person's details.
Agents can use a database or internal tool without anything of yours being opened to the internet. One small piece runs inside your network and agents reach through it.
Where the AI runs
With Ollama, an AI model runs on a computer inside your network. The agent sends its questions there instead of to an AI company.
The model itself, and what the agent asks it and what it answers while it works. One small piece in your network connects out to Monsel, so nothing is opened to the internet.
Your documents in Studio, your agents' instructions and the run records. A few supporting steps can still use a hosted AI: guardrail checks, and rewording a chat reply before it is sent.
Open models vary in quality and in speed, which depends on your hardware. Try your job on one before you rely on it. This is on the Pro plan and above. How it works
Agents from the marketplace
Agents you install from other people run in your workspace under your limits. How the marketplace works
An installed agent can be given only the projects you hold yourself. It arrives paused, so you read it before it runs.
Every listing has a summary of what it does and what it can reach, written by Monsel and not by the seller. Listings that have not been certified carry a warning that Monsel has not verified them.
Working with other agents
An agent hands work to another only where you have allowed that pairing, inside your workspace and across company lines.
A partner's agents sign in with logins you create for them. Cancel one and the access stops from that moment.
A published standard lets your agents work with a partner's, even on a different platform. How that works.
When both agents run on Monsel, a call from one to the other appears in the run record at both ends: who asked, who answered, and on whose behalf.
Ownership
Your documents, skills and the record of what agents did can be exported on request. If you change AI company, or leave, you do not start again from an empty page.
An organisation can have a separate copy of Monsel. Tell us what your policy requires, including where data must be kept, and we will say what we can do.
What we do not claim yet
Private beta
Monsel is in private beta with a small number of businesses. Leave an email and we will tell you when the next batch opens. 200 free credits come with your first sign-up.
To talk first, ask us to set agents up for your business from a written proposal, or write to hello@monsel.ai.
Analytics
With your permission, we use privacy-friendly analytics to see which pages are read. It sets no cookies and does not follow you across sites. We use no advertising and no cross-site tracking. Privacy policy