RacterSentry monitors, gates, and audits every action your AI coding agents take — in real time. Endpoint detection and response, purpose-built for the age of autonomous code.
$ racter-sentry init --token rs_live_...
✓ Agent registered (mTLS certificate issued)
✓ Claude Code hooks configured
✓ OTel telemetry routed to localhost:4317
✓ eBPF kernel module loaded
✓ Sigma rules synced (47 rules, v2.1.0)
Sentry is active. Monitoring 1 agent session.AI coding agents operate with broad permissions and zero security awareness. Every session is an unmonitored attack surface.
AI agents can run rm -rf /, DROP TABLE, or chmod 777 without understanding the consequences.
Agents hardcode API keys, database passwords, and private keys directly into source files — then commit them.
Agents install unknown packages that could be typosquatting or malicious. One bad dependency compromises the entire build.
Three steps from install to full coverage. No configuration drift. No maintenance burden.
The agent installs as a systemd service, loads eBPF into the kernel, and configures Claude Code hooks automatically. Under two minutes from download to active monitoring.
Sigma rules evaluate every tool call in real time. Block destructive commands, flag secrets, catch scope violations — before the code is written.
Findings stream to your dashboard. Acknowledge, investigate, or auto-route to Slack, PagerDuty, or your SIEM. Full audit trail for compliance.
Industry-standard Sigma rules. Write your own or use our curated library.
title: Destructive Bash Command
id: atr-001
status: stable
level: critical
description: Detects destructive commands executed by AI agents
tags:
- attack.impact
- owasp.agentic.asi02
detection:
selection:
tool_name: Bash
keywords:
bash_command|contains:
- 'rm -rf'
- 'mkfs'
- 'dd if='
- 'chmod 777'
condition: selection and keywords
action: blockSix detection categories covering the full spectrum of AI agent risk.
Block file system destruction, database drops, and permission changes before they execute.
Detect API keys, private keys, and passwords written into source files or passed through tool calls.
Alert when agents access files outside the project workspace, including SSH keys, AWS credentials, and system configs.
Flag unknown package installations and check for typosquatting, low download counts, and recently published packages.
Monitor for curl, wget, and network calls that could exfiltrate source code, secrets, or sensitive data to external endpoints.
Track permission mode changes, auto-approve toggling, and subagent delegation that expands the agent's authority.
Built for security teams deploying AI agents at scale.
Kernel-level protection on Linux. Detects and reports attempts to kill, debug, or modify the agent process.
No static tokens. Short-lived certificates rotate every 24 hours. Compromised credentials expire automatically.
Every detection rule maps to ASI01 through ASI10. Full coverage of the OWASP Agentic Security standard.
Source code, prompts, and file contents never leave the developer's machine. Only structured findings reach the SaaS platform.
One price per agent. Tiers differentiated by feature set, not cost.
Coverage across the AI coding agent landscape.
| Agent | Coverage | Capabilities |
|---|---|---|
| Claude Code | Full | Gate + Observe + Transform |
| Kiro | Full | Gate + Observe + Transform |
| Cursor | Strong | Observe + Transform |
| GitHub Copilot | Moderate | Observe |
| Windsurf | Moderate | Observe |
| Codex CLI | Moderate | Observe |