Connect approved tools to your agents, with fine-grained control
Minerva works with the apps your team already uses. IT decides which ones agents may use, and down to which folder, channel, label or customer.
The secure agent for work
Minerva is the agent you can finally trust with your company data. Thanks to strong isolation and tight access control, Minerva is the first solution for connecting powerful AI to the software your business relies on.
Let employees move faster with agents. Keep IT in control.
- GmailLabels Customers and Receipts
- Slack#support and #acme-account
- XeroNot approved for agents
Also available- Notion
- Google Drive
- Microsoft Teams
- Jira
- HubSpot
- GitHub
- +13
Give AI the context of your organisation
Connect your agents to knowledge bases like Notion and Confluence, and to email like Gmail. They learn how your company works, from only the pages, labels and channels you choose.
Secure your agents and control what they do
Each agent runs isolated and never holds a key. Every action is checked against your rules before it reaches an app, so even a tricked agent cannot go further than you allowed.
How Minerva compares with other agent products
Minerva is younger than these products and connects to fewer apps. It differs in how access is granted and enforced.
| Minerva | OpenClaw | Hermes Agent | OpenAI dots | Grok Bot | |
|---|---|---|---|---|---|
| The agent never holds your keys | Yes | Kept on your machine | Kept on your machine | Kept by OpenAI | Signs in as you |
| Access limited per folder, channel or customer | Yes | Per tool | Per tool | Automatic review | Asks before acting |
| Each task runs in its own sandbox | Yes | On your machine | Optional | Yes | One shared machine |
| Choose your model | Planned | Yes | Yes | OpenAI only | Grok only |
| Scheduled tasks | Planned | Yes | Yes | Yes | Yes |
| Memory across conversations | Planned | Yes | Yes | Yes | Yes |
| Integrations | 20 | Many, via skills | Many, via MCP | 4,000+ | Any site in a browser |
Who is building Minerva
I’m Szymon Nastaly, and I build AI agents for code review and security testing. One is a pull-request review agent used by more than 160 developers, another a multi-agent system for penetration testing. I’m studying computer science at ETH Zürich, with a focus on security. I started Minerva because I wanted agents to work in company systems without handing them the keys.
I write about what I learn from building with agents in This Week in Agents.
Supported by
- Websiteszymonnastaly.com
- LinkedInin/szymon-nastaly
- GitHubSzymonNastaly
- NewsletterThis Week in Agents
FAQ
Is Minerva open source?
Its engine is. The engine runs the agents and includes the gateway, the sandbox and every connector. Its source is on GitHub under the MIT licence. The web app around it, with accounts, organisations and billing, is not published. We offer Minerva only as a hosted service and do not support running the engine yourself.
Which models can I use?
The one we choose for the hosted service. We call it with our own key, so your company needs no account or key with a model provider. Choosing another model is planned.
How is this different from approving each action?
Approval prompts rely on someone reading every request, and the agent keeps the keys while it waits. In Minerva you set the rules ahead of time and the gateway refuses anything outside them; approval for single actions is planned as an addition.
Can agents run on a schedule or be reached from Slack and Teams?
Not yet. Both come after memory across conversations, which we are building next, and they will go through the same gateway and checks as a conversation.
Can I connect MCP servers or install community skills?
Not at present. A generic MCP server does not tell the gateway which folder, channel or customer a call will touch, so it could only be allowed or denied as a whole.
Where are my credentials stored?
Encrypted at rest in PostgreSQL with versioned keys, readable only by the backend. They never reach the agent’s container or the browser.
Is it only for companies?
Yes. Every account belongs to a company. We set up each company in Minerva after talking with them, and the company’s admins then invite the employees who need an account. Nobody can sign up on their own.
Is it ready for production use?
Not yet. The hosted service has not launched, and it opens to a few companies first, as pilots. The permissions, the sandbox and the connectors already work against live accounts. Admins can invite employees and approve tools, and employees can share their agents. Memory across conversations, and a page where a company’s admins read its audit log, are not built yet.
What will Minerva cost?
We have not published pricing yet. People who asked for early access will hear about it first.
Get early access to Minerva
Minerva opens to a few companies first. Tell us about yours in a minute, with three quick taps, and we will write to you about a pilot.
The source of Minerva’s engine is on GitHub, for you or your security team to review.
We ask about
- The size of your company
- What you would hand to Minerva first
- When you could start a pilot