How is Oscar delivered?
Oscar is delivered as an MCP server. You connect Oscar from inside the LLM application you already use, such as Claude, Claude Code, or ChatGPT, and any harness that supports MCP servers works. Operator Edition runs the server locally on your workstation, Oscar Pro is hosted by OpsCanvas, and Enterprise is self-hosted in your own environment.
Which models can I use with Oscar?
Any model. Oscar works with the LLM you already use in every edition, and Enterprise also supports self-hosted models running entirely on your network.
What makes Oscar safe to use on my cloud?
Oscar is read-only by default, limited to the credentials you already have, and grounded in topology and blast radius context so it understands what an action touches before proposing it. Execution is deterministic and approval-gated: nothing in your environment changes without you.
What is the adaptive interface?
Oscar works in standard text chat, and when a task calls for it a graphical interface appears in harnesses that support it. This experience is built on the MCP Apps standard.
What are Baseline Assessments?
Baseline Assessments establish a trusted operational baseline of your environment and are included in every edition. Pro and Customized Assessments, which go deeper into specific domains, are provided and administered by OpsCanvas Partners.
Does OpsCanvas see my cloud data on Oscar Pro?
No. Oscar Pro is the only OpsCanvas-hosted edition, and OpsCanvas receives only usage and performance metrics, never your cloud data or credentials.
What is included in OpsCanvas Enterprise?
Enterprise is the fully self-hosted edition: everything in Oscar Pro, an org-wide topology and context graph, support for self-hosted models, and your data staying on your network. Additional enterprise governance capabilities are rolling out through early access; contact sales for the current list.
When is Oscar generally available?
Oscar is in early access now, and v1 ships this September. You can start free with Operator Edition today.