Start protecting AI in an afternoon. No firewall ticket.
Three deployment paths. Pick one, or run them side by side.
Choose a path and start enforcing today.
- 01
SELECT
Choose PAC file, Browser Shield or API proxy based on where your AI traffic originates.
- 02
CONFIGURE
Set policies, risk thresholds and detector packs in the console — no code required.
- 03
DEPLOY
Push a file, install an extension or change one base URL. No firewall change window and no agent rollout — each path is reversible.
- 04
EVIDENCE
Verdicts and decisions flow into the tamper-evident audit chain from day one.
What it does.
PAC file deployment
Generate a proxy auto-config file and push it via Windows GPO or macOS MDM.
Browser Shield extension
Manifest V3 extension for Chrome, Edge, Brave and Arc with local prompt inspection.
OpenAI-compatible API proxy
Change one base URL and every SDK call is governed without model or vendor lock-in.
Mixed-mode operation
Run PAC, Browser Shield and API proxy side by side for layered coverage.
Generated policy artifacts
.reg and .mobileconfig files are created automatically for managed rollout.
Same-day evidence
Audit-chain entries start accumulating the moment the first prompt is processed.
One line of code.
policy > "Switch the OpenAI SDK base URL to route every completion through XAIGuard"
client: OpenAI(base_url="https://api.xaiguard.com/v1")
model: "gpt-4o"
policy: "block-pii-and-secrets"
prompt_risk: HIGH — PII detected
verdict: "BLOCK_AND_REDACT"
audit_hash: "ed25519:7d1a…e03b"
Deployment is designed to be low-risk and reversible, not a six-month network project.
Three ways in. One policy everywhere.
PAC file
- Generate a proxy auto-config file from the console.
- Push it with Windows GPO or macOS MDM.
- .reg and .mobileconfig files are generated for you.
- AI traffic routes through XAIGuard enforcement.
Good for: full org coverage without touching the network edge.
Browser Shield
- Install the Manifest V3 extension for Chrome, Edge, Brave or Arc.
- Prompts are inspected locally before submission.
- Six built-in detectors plus your custom patterns.
- Blocks are reported to your tenant and written to the audit chain.
Good for: fast pilots and BYOD.
Chrome Web Store listing is in review. The signed package is available for managed install today.
API proxy
- Use the OpenAI-compatible endpoint.
- Change one base URL in your SDK or agent code.
- Every model call is scored, redacted or blocked.
- Works with any model that speaks the OpenAI chat format.
Good for: your own applications and agents.
Pick the fit for your environment.
| Criteria | PAC file | Browser Shield | API proxy |
|---|---|---|---|
| Coverage | All browser AI traffic | Browser-based assistants | Your apps and agents |
| Setup time | Hours | Minutes | Minutes |
| Network change required | No | No | No |
| Works off-network | No | Yes | Yes |
| Best for | Full org rollout | Fast pilots / BYOD | Custom applications |
From pilot to evidence in three weeks.
Pilot group
Deploy Browser Shield or API proxy to a small team and run first prompts.
Policy tuning
Tune thresholds, add custom detectors and review false positives.
Org-wide
Roll out PAC file or push extension to all users and agents.
First evidence export
Generate a signed evidence package for auditors or board review.