Know where Claude runs
CodeSensor records the foundation models each repository calls and standardizes them by family, Claude included. With your SIEM connected, Shadow AI shows which AI services your people use and how much data moves through them.
Claude powers your products and your developers’ agents. Cranium ONE inspects every prompt and response, enforces your policy in line, and catches poisoned instructions before an agent ever reads them.
CodeSensor records the foundation models each repository calls and standardizes them by family, Claude included. With your SIEM connected, Shadow AI shows which AI services your people use and how much data moves through them.
Guardian inspects the traffic between your application and Claude. Sessions, events and the signals each one triggered land in Monitoring, with nothing in the request path in Listen mode.
Profiles decide which signals matter and what happens to each: block, modify or pass. Different event types — user to agent, agent to agent — can carry different profiles.
The free Adversarial Inputs Detector for VS Code and JetBrains scans agent files — CLAUDE.md, AGENTS.md, .cursor, .windsurf, .github — for malicious instructions, exfiltration endpoints and hidden Unicode. Enforce mode guards Claude in line at runtime.
Trace shows why each decision was made, from deterministic classifiers rather than another model’s opinion, and AI Cards carry the evidence to customers and auditors.
Cranium ONE works with the stack you already run. Your policy stays yours — we make it enforceable.
In Guardian, add a Claude model with its vendor model ID and your Anthropic API key.
Route an application through Guardian in Listen mode and read the sessions in Monitoring.
Install the Adversarial Inputs Detector so agent instruction files are checked before anyone clones and runs.
You add a Claude model in Guardian with its vendor model ID and your Anthropic API key. Guardian then evaluates the prompts and responses that pass between your application and the model.
Yes. The free Adversarial Inputs Detector plugins for VS Code and JetBrains scan agent instruction files such as CLAUDE.md, AGENTS.md and GEMINI.md, plus .cursor, .windsurf and .github folders, for malicious instructions, exfiltration endpoints and hidden Unicode.
No. Guardian’s verdicts come from deterministic, purpose-built classifiers, and Trace explains each one. That keeps decisions consistent and explainable.
Yes. Listen mode evaluates a copy of the traffic in parallel and records what it would have done. Enforce mode is what lets Guardian block, modify or pass prompts and responses in line.
We use essential cookies to run this site, and — only with your permission — analytics and marketing cookies to improve it. You're the hero here; you stay in control. Privacy Policy.
Amazon Bedrock
OpenAI
Microsoft Azure