Autonomous agents
Business processes, run with AI
VossMoos engineers and operates autonomous agents for customer service, product marketing, sales development and robotic compliance. We engineer the whole solution: the systems the agents run on, the connections to the tools you already use, and their day-to-day operation.
Mossy · open-source agent runtime
Chats are stored so our engineers can follow up. Details in our privacy policy.
Let your team do ten times more work with the same effort
Every operation faster than now. We start from a problem you already have, not from a technology. The agents work in the systems your team already uses, within the limits you set, and your people make the decisions where you require it.
One team, from the first analysis to the running system
An agent needs more than a model. It needs connections to the systems you use, infrastructure to run on, monitoring, and someone responsible when something changes. We engineer and operate all of it, and we run our own work the same way.
- Process design: how the agent should work, and where a person decides.
- Software and AI: the logic, models and company knowledge the agent needs.
- Connections: to your ERP, CRM, help desk, documents and email.
- Infrastructure: hosted by us, in your cloud or on your own servers.
- Operation: monitoring, updates and improvements after launch.
Open infrastructure for autonomous agents
We build agents that work in real companies, our own included. What we learned is now open-source infrastructure: an IAM layer for AI agents, an agent runtime and a corporate AI knowledge layer. You run it on your own cloud or servers.
- Reliable: proven in live processes, not in a lab.
- Verifiable: every line of code is open, and every access decision is logged.
- No lock-in: open-source licences and any AI model. You can run it without us.
-
01 IAM layer for AI agentsMyco
Who may see and do what.
One endpoint per agent with a reviewed set of tools for your SaaS apps. Each agent sees only the tools and data its roles allow, and every decision is logged.
- Roles per agent and per user, explicit deny wins
- Rows scoped and sensitive fields redacted
- Permissions as YAML in Git
-
02 AI agent runtime
Mossy
The agent runtime.
A skill-first runtime that runs AI-operated work in production. New capabilities are added as skills, and work moves through channels and queues.
- Composable workers and skills
- Channels and queues
- Reaches company tools through the IAM layer
-
03 Corporate AI knowledge layerEtchMem
Memory that reconciles itself.
What one process learns or a colleague corrects becomes knowledge every agent can use. EtchMem keeps one current answer per fact and flags where systems disagree.
- Learns from work and from corrections
- Facts reconciled across systems
- A date and source for every answer
See your problem solved before you commit
Tell us one process that costs your team time. We show you, free of charge, how the three layers would handle it: which data the agent reaches, what it does, and what it remembers.
- 1You describe the problem in a short call.
- 2We prepare a demo of the three layers on your case.
- 3We walk you through it, and you decide what happens next.
People stay in control
Each agent acts within the limits you set, refers uncertain cases to a person and keeps a complete record of its work.
- Human approval where you require it, set by task type, value or confidence.
- A complete audit trail of every input, decision and action.
- Limited permissions: each agent reaches only the data its task requires.
- Your choice of AI model, commercial or open-weight.
- OpenAI
- Anthropic
- GLM
- Model Context Protocol
- FastMCP
- Python
- Pydantic AI
- FastAPI
- PostgreSQL
- DuckDB
- Docker
- Kubernetes