Reliable, secure, governed agents
Build agents you can rely on to do the work.
An agent is a colleague you set up for one job: raising invoices, building quotes, sorting the inbox, answering customers. It works from your documents and uses your tools. You don't have to know how to write code. Work you choose to check waits for you: approve it, edit it, or send it back with a reason.
Private beta, free to start. See how it works.
What you are building on
Agents that are reliable, secure and governed from the first day.
Reliable
Totals on invoices and quotes are worked out by the platform, not guessed by the AI. Every step reports passed or failed, and a failed step is never reported as done.
Secure
An agent can use only the tools and documents you give it. Personal details are masked before a conversation is stored.
Governed
Work you mark for approval waits for a person. Who approved what, when and why is kept, and every run can be read back.
Built for the person who knows the job
You don't have to know how to write code.
If you can explain the job to a new colleague, you can build the agent.
Write instructions as sentences
Say what the agent should do, what it must never do and when it should ask you first.
Connect from a list
Pick email, WhatsApp, Telegram, your calendar or Google Drive and switch it on.
Read what it did, in plain language
Every run shows what the agent read, what it used and how it ended.
From a folder of documents to an agent doing a job
Five steps, and you choose which work waits for your approval.
- Capture Teach it your business
- Design Write instructions as sentences
- Connect Phone, inbox, calendar, forms
- Approve You check the work you choose
- Improve Read the runs, refine the instructions
Built to finish the job
An agent carries a job through to the end.
An agent builds the invoice, drafts the quote, sorts the inbox and sends the briefing. Monsel is built around that.
- Something starts it A message, an email, a schedule
- It does the work Reads your documents, uses your tools
- Each step reports pass or fail A failed step is reported as failed
- You approve what you choose Before it goes out
- It is on record Every run can be read back
It uses your tools
Send the email, build the PDF, look up the order. An agent has only the tools you give it.
Sums are not guessed
With a template, the totals on an invoice or quote are worked out by the platform, not by the AI.
Failures are not hidden
If a step fails, the run says so. A reply that could not be delivered is not counted as done.
It carries on without you
It can wait for your approval or for another agent, then pick up where it left off.
The AI can still make mistakes. That is why you choose which work waits for you, and why every run is kept on record.
Fewer guesses
The platform handles what does not need judgement.
The model is asked to think only when thinking is needed. Monsel does the rest, the same way every time.
Sums are done by code
Invoice totals are calculated, not estimated.
Search needs no AI model
It finds the documents. The model writes the answer.
Every step is checked
Passed, needs action or failed. A failed step is reported as failed.
Agents see only what you gave them
An agent can use only the tools you grant it.
Two products, one account
Studio holds what your business knows. The Agent Orchestrator runs the agents that do the work.
Update a policy once and every agent works from the new version from the next job.
The memory your agents work from.
- Agents read from it before they raise an invoice or draft a reply.
- Documents and PDFs become searchable without anyone filing them.
- Every answer shows the document it came from.
Colleagues you brief in plain language, with a person where you choose.
- Brief an agent the way you would brief a colleague. No code.
- Start it from webchat, WhatsApp, Telegram, email or a schedule.
Approval
Work you mark for approval waits for a person.
A customer reply, an invoice, a plan for another agent. The person sees the documents the agent read and the tools it used, on the same screen.
Who approved what, when and why is kept on the document.
Live handoff
The agent can hand a conversation to a colleague.
A customer is upset, asks for a person, or asks something outside the agent's job. The agent moves the conversation to a shared queue and says so. A colleague picks it up on the same channel, with the whole conversation in front of them, and can hand it back.
One queue for every channel
Webchat, Telegram and WhatsApp conversations land in one place. Whoever is on shift claims one and replies as themselves.
Reach you where you are
A chime and a badge on screen, and a push notification on your phone.
Tools and AI
What it connects to, and what it costs.
The tools you already use.
Email, WhatsApp, Telegram, webchat, calendars and Google Drive are built in. Dropbox and OneDrive are coming. Other business tools can be added. Tools inside your own network stay there, with nothing opened to the internet.
Pay for the AI your way.
Use an AI subscription or account you already have, or let us supply the AI. The plan price is the same either way, and it is not charged per message. Free to start, with 200 free credits at first sign-up.
How pricing worksOr keep the AI on your own computer.
Run an AI model with Ollama on a computer in your own network. Agents still work in Monsel, with your documents, approvals and records, but their thinking happens on your machine instead of at an AI company. A few supporting steps, such as guardrail checks, still use a hosted AI, and it is on the Pro plan and above. Open models vary in quality, so try your job on one before you rely on it.
- The agent works in Monsel With your documents and approvals
- One small piece in your network Connects out, no ports opened
- An AI model on your computer Run with Ollama
- The answer comes back The job carries on as normal
Start from an agent that already works.
The marketplace has agents other people have built. Certified agents are run on real test cases, with the results shown. Before you install one, you can read what it does, what it needs from you and what a run costs. It arrives paused, and it can only reach the projects you give it.
See the marketplaceUse cases
What businesses put an agent on first.
Back office
Invoices, briefings and inbox sorting, overnight.
Sales and quotes
Enquiries answered out of hours, quotes held for you.
Customer support
Answers from your own policies on webchat, WhatsApp and email.
Ask the company
Staff questions answered from the handbook and policies.
For yourself
A personal agent on Telegram or WhatsApp that handles the small jobs and tells you when it needs you.
For the technical and the curious
The technical detail.
How connections, access codes and permissions between agents work, with the proper names for each.
Private beta
Workspaces open in batches.
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.