Features
What a Sentisec workspace gives you — drop-in compatibility with the clients you already use, key management, and the controls that decide what gets enforced.
Three things make up the product surface you interact with: how your agents connect, the keys involved, and what you control.
Drop-in replacement
Change one base URL. Your Anthropic or OpenAI client keeps working unchanged.
Keys management
The workspace API keys your agents use, and the provider credentials Sentisec uses on your behalf.
Workspace controls
Protection setup, the posture ladder, the egress allowlist, alerts and SIEM delivery.
How it fits together
You point an agent at a workspace-scoped base URL with a workspace API key. Sentisec resolves the workspace from that key, applies that workspace's settings, forwards the request to the upstream your provider credential unlocks, and evaluates every tool call the model proposes along the way.
Whether a flagged call is actually stopped comes down to one setting — your workspace's rung on the posture ladder. On the two lower rungs you get the full record and no behaviour change; on the two upper ones the call is neutralised before it runs. Every workspace starts on the lowest, and moving up takes evidence rather than a toggle.
The egress allowlist is the one thing worth configuring early: declaring where your agents legitimately reach is what makes a call aimed anywhere else stand out.
Related
- Quickstart — first monitored call
- Dashboard — the operator surface for all of it
- Enterprise — SSO, RBAC, SCIM, in-VPC