CLOSED BETA

Defensive Firewall
for AI Agents

ClawGuard intercepts and sanitizes inbound email before it ever reaches your LLM. Prompt injection, hidden scripts, secret exfiltration—neutralized automatically.

clawguard — inbound scan

Six layers of protection

Every email passes through deterministic, regex-based analysis before your agent sees a single token. No AI in the loop—pure, auditable rules.

Injection Detection

Catches prompt injection, jailbreak attempts, role-override patterns, and instruction hijacking with 40+ detection rules.

HTML & Script Stripping

Removes all HTML, CSS, JavaScript, hidden elements, tracking pixels, and invisible text. Your agent sees only clean plaintext.

Secret Redaction

Detects and redacts API keys, passwords, JWTs, AWS credentials, and other secrets before they leak into agent context.

Attachment Filtering

Allowlist-based attachment policy. Blocks executable payloads, extracts safe text, reports blocked files with metadata.

Risk Scoring

Every email receives a 0–100 risk score with granular flags: injection, HTML, secrets, Unicode abuse, hidden content.

Real-Time Monitoring

Live dashboard with event stream, timeline graphs, sender analytics, and per-inbox filtering. See threats as they arrive.

Three steps to safe email

Connect your inbox, let ClawGuard process every message, and pipe only sanitized payloads to your agent.

1

Connect Your Email

Authorize your Gmail directly from the dashboard. No SSH, no config files. OAuth in two clicks.

2

Automatic Sanitization

Every inbound email is stripped, scanned, and scored. Injection patterns are flagged and redacted deterministically.

3

Safe for Your Agent

Clean, structured JSON payloads are forwarded to your agent or accessed via API key. Risk metadata included.

Request Early Access

ClawGuard is in closed beta. We're onboarding teams building LLM agents that process external email. Reach out for an invite.

Request Access