Products / AI Cloud
sovereign AI · AI Cloud
AI Cloud: private AI in your own cloud account
A chat workspace, open-weight models and connections to your internal data, inside a cloud account or Kubernetes cluster you control. Your staff sign in through your identity provider, and Clover Tech keeps no credentials or access after install.
deployment
What runs where
Workspace
Chat, projects and agents in the browser
Models
Two open-weight models, plus Claude or GPT if you elect them
RAG connections
Three internal data sources, more as add-ons
Identity and audit
Your single sign-on and roles, with audit logs kept here
Prepares, tests and signs each upgrade. Your team applies it.
Bills compute, GPU and storage to your account.
Hosted by Clover Tech until the next release moves it into your account.
the seat
Every seat includes
- Single sign-on through SAML or OIDC, SCIM provisioning, roles and audit logs, all kept in your account
- Controls documentation aligned to SOC 2
- Zero data retention at Clover Tech, and conversation retention set by your admins
- Spend controls: per-user metering, monthly and daily caps and a per-agent kill switch in development
- The workspace, with projects that switch between models per chat
- Agents with instructions, file search and code execution
- Three RAG connections, each to one internal data source
- Two open-weight models, plus Claude or GPT on your own vendor key with no Clover Tech fee
- Automatic routing of routine work to open-weight models and harder tasks to frontier models in development
- Standard support, covering installation, production issues and model updates
- A usage dashboard that reports by team, with a minimum group size in development
5-seat minimum, with no setup or platform fee. Your cloud provider bills compute separately. The price list has the add-ons and example bills.
add-on products
Add products on the same deployment
Clover Code
availableVS Code in the browser, connected to an open-weight coding model in your environment.
MCP server
availableLets an AI tool such as Claude Desktop or Cursor deploy Clover Tech Marketplace apps in your account. Clover Tech never holds its keys.
Agent payments
in developmentAgents pay in USDC on Base or Solana, or Bitcoin over Lightning, within limits you set. Keys stay in your account.
specifications
Specifications
| Deployment | Your cloud account, on a VM, a container host or your own Kubernetes cluster. The installer needs a bare machine with SSH. |
|---|---|
| Regions | Every provider in the install wizard, in the large majority of each provider’s regions, EU included. |
| Identity | Single sign-on through SAML or OIDC is required for every user, with SCIM provisioning and roles. |
| Audit and logs | Audit logs stay in your account, and logs forward to your MSSP. |
| Vendor access | None after install. Clover Tech holds no credentials, support works from operational logs and monitoring, and your team applies each upgrade Clover Tech signs. |
| Models | Two open-weight models from families such as Llama, Qwen, Mistral and Gemma, served by vLLM or Ollama. Claude and GPT run on your own vendor key, and that traffic goes to the vendor. |
| Data flow | Documents, conversations and audit logs stay in your account. Prompts to the bundled models pass through a model gateway Clover Tech hosts until the next release moves it into your account. Clover Tech receives operational logs and monitoring results only. Claude and GPT traffic on your own vendor key goes to that vendor. |
| Retention | Zero data retention at Clover Tech. The model gateway writes no prompt or completion to any log, trace, crash report or cache, and operational logs carry no prompt or completion content. Your admins set how long conversations are kept in your account. Claude and GPT traffic follows your agreement with that vendor. |
| Upgrades | Clover Tech prepares, tests and signs each upgrade. Your team applies it. |
| Compliance | Controls documentation aligned to SOC 2 in every deployment. Mapping to another framework, starting with GDPR, is an add-on. |
| Support | Standard support is included. Priority and 24 × 7 levels are add-ons, and every level commits to a response time. Your cloud provider runs the compute and network, so there is no uptime SLA. |
RAG connections
RAG connection catalog
Three connections come with every deployment, and each additional data source is $1,000 a year. Connectors become available as each one is certified, so ask which are certified today.
SharePoint, Google Drive, Box, Dropbox, Confluence, Notion
Salesforce, HubSpot, Dynamics, Affinity
Snowflake, BigQuery, Databricks, Postgres, S3
Splunk, Elastic, Sentinel, Wazuh, CrowdStrike
Slack, Teams, Outlook and Exchange, Gmail
Jira, ServiceNow, Zendesk, Linear
NetSuite, QuickBooks, SAP, Workday
GitHub, GitLab
roadmap
In development
Ask us for current timing on these and on the items marked in development elsewhere on this site.
- Model gateway inside your account, in the next release
- Claude and other frontier models through Amazon Bedrock in your account
- Shared team projects
- Audit feed from your account into your SIEM
- Agent registry listing every agent and app that is running
- A Slack app, then Microsoft Teams
- An EU region for each model
Why not use the AI team plan we already pay for?
On a team plan, every prompt goes to the vendor, and retention and export controls usually sit on enterprise tiers. In AI Cloud, conversations and audit logs stay in your account, kept as long as your admins set. The seat matches a Claude Team or ChatGPT Business seat at $20 a month billed annually. Those plans include model usage, while AI Cloud runs on compute you pay your provider for.
What access does Clover Tech have to our environment?
None after install. Clover Tech holds no credentials, support works from operational logs and monitoring, and your team applies each upgrade Clover Tech signs. Single sign-on through SAML or OIDC is required for every user, and audit logs stay in your account and forward to your MSSP.
Who installs and maintains it?
Your team runs the setup wizard, and Standard support covers installation, so no ML platform team is needed. Clover Tech prepares, tests and signs each upgrade, and your team applies it.
Are open-weight models as good as Claude or GPT?
Not quite. Epoch AI measured open-weight models trailing the best closed models by about four months on average in 2026. Claude and GPT stay available when you elect them, and you choose which models your staff can use.
Does any of our data reach Clover Tech?
Prompts to the bundled models pass through a model gateway Clover Tech hosts today, and the next release moves that gateway into your account. The gateway runs under zero data retention (ZDR): it writes no prompt or completion to any log, trace, crash report or cache. Documents, conversations and audit logs stay in your account, and your admins set how long conversations are kept. Clover Tech receives operational logs and monitoring results, which carry no prompt or completion content, and has no access to your cloud account after install. Prompts you send to Claude or GPT on your own vendor key go to that vendor, and your agreement with Anthropic or OpenAI sets how long that vendor keeps them, including any zero data retention terms you sign with them.
Is AI Cloud SOC 2 certified?
Every deployment includes controls documentation aligned to SOC 2. The security report add-on ($1,200 a year) builds a quarterly deployment report and an annual SOC 2 evidence pack from your audit logs. Mapping to another framework, starting with GDPR, is $15,000 a year per framework. Our SOC 2 Type II audit is still scheduled for the first quarter of 2027.
What does compute cost?
Your cloud provider bills it to your account. One NVIDIA L40S on AWS serves both included models for up to about 50 users, an estimated $12,300 to $16,300 a year. The price list estimates compute for each example configuration.
Do our staff need a new chat tool?
Today they use the workspace in a browser. A Slack app, then Microsoft Teams, is in development.
next step
Book a 30-minute conversation
Tell us your user count, your cloud provider and the data sources you want connected. We reply by email to set a time.