No SDK required. No configuration. One HTTP call.
Register at the dashboard. Name your agent, get an API key. Takes 30 seconds.
One HTTP POST from your agent. Any language. No SDK, no setup, no install.
Every action streams to your dashboard in real-time. Set alerts. Export logs.
Live event stream, every decision, trade, and error, as it happens
SIEM-grade security for AI agents. Financial monitoring built-in, not bolted on.
Every agent action is captured, normalized to SIEM standards (CRITICAL/HIGH/MEDIUM/LOW/INFO), and stored with a full audit trail. CEF + LEEF export to Splunk, Datadog, QRadar.
Live activity timeline, decisions, tool calls, errors, and anomalies stream to your dashboard the moment they happen. Like a SOC console, built for agents.
Events are automatically tagged to SOC 2, PCI DSS, ISO 27001, NIST CSF, and GDPR controls. Generate audit-ready compliance reports in one click.
Silence detection, error-rate thresholds, spend limits. Related events are auto-grouped into incidents. Email and webhook delivery.
Unlike generic SIEMs, VIGIL understands crypto, SOL cost per action, P&L tracking, on-chain transaction feed. Financial monitoring built-in, not bolted on.
Logs live in your account only. Export everything as JSON or CSV anytime. Zero lock-in. COSTRINITY never accesses, shares, or sells your agent data. Ever.
One API call. Works with OpenClaw, Claude, GPT, AutoGPT, ElizaOS, or any custom script.
npm install @costrinity/vigilπ Any language. REST API works from Python, Go, Rust, bash, anything that can make an HTTP call.
VIGIL is the developer-grade SIEM for AI agents. Free, local-first, and built to feed enterprise security stacks via CEF/LEEF export when you need them.
Different lanes
GuardionAI, Silverfort, and SentinelOne are excellent at what they do. They're also enterprise-priced and aimed at CISO-level deployments. VIGIL is the developer-grade layer that sits beneath them, and feeds into them via CEF/LEEF when the time comes.
The pitch: if you're building agents, start with VIGIL, it's free, local-first, and gives you SIEM-grade output day one. When you scale to a CISO-grade stack, VIGIL feeds into it. You don't replace one with the other; you compose.
Recent updates
What's shipped recently
Activity Graph (force-directed agent network) Β· Local-first by default Β· Desktop app no-login + auto-key Β· Auto-update channel Β· Mobile rewrite
ElizaOS plugin v2 Β· batch ingest Β· webhook retry logic Β· P&L export
P&L dashboard Β· SOL cost tracking per action Β· 7-day profit charts Β· on-chain tx feed
Smart alerts Β· email notifications Β· agent silence detection Β· multi-agent filter
The Activity Graph clusters every agent event as a live force-directed network: agents in the center, threats VIGIL caught around them. Red nodes are blocked: SQL DROP TABLE, shell rm -rf, secret leaks, exfiltration attempts. Orange is high severity (prompt injection, wallet drain). Yellow is anomaly (unusual tool-call rate). Green is fine. Click any node for the full payload, severity, cost, and compliance tags.
Start free. Scale when you need to.
COSTRINITY does not store, share, sell, or access your agent event data for any purpose. Your logs are isolated to your account, we can't read them even if we wanted to. Export everything as JSON or CSV and delete your account at any time.
Free forever for 1 agent. No credit card. Up and running in 90 seconds.
Free forever for 1 agent Β· No credit card required