Secure LLM traffic
Check-points and behavioral signals on every model call. Route through {provider}.securellm.burgus.ai — keep your provider keys and SDK.
Burgus Security · production AI
Secure production AI — every model call and data connection
Burgus Security is a production security layer for agent workloads: LLM API traffic, MCP tool and data connections, and collaboration across multi-agent systems. Check-points, behavioral learning, access control, and audit — alert-only, built for high-volume production.
Because Burgus secures at the connection — independently of any single cloud or model vendor — you keep choosing the models and MCP sources that fit your work. Same SDKs and keys (BYOK), direct provider relationships, and one predictable security subscription.
LLM + MCP security
Multi-agent
4,000+ check-points
BYOK · your providers
USD 189/mo security
burgus — a Roman watchpoint on high ground. Observe, signal, stand guard.
What Burgus secures
Security for every connection your agents use
Burgus Security sits between your agents and the models and data they reach — not to replace your providers, but to secure, audit, and govern production traffic. That architecture is what lets you route any supported LLM or MCP source without adopting a bundled vendor stack.
Secure MCP connections
Audit and govern tool ingress and egress through *.mcpdata.burgus.ai. Access control per source and agent.
Your choice of providers
Security is separate from inference — so you pick Groq, OpenAI, Anthropic, or any supported model on your terms. USD 189/mo for the security layer.
Why Burgus
Serious AI needs serious security — and room to choose your stack
Production teams run diverse models and data sources because different jobs need different tools. What they need first is security across all of it — who called which model, which agent reached which data source, which alerts fired as workloads scale. Burgus delivers that security story on live production traffic.
Burgus Security is the purpose — securing LLM and MCP production and multi-agent collaboration. Choosing your providers freely is what becomes possible when security is not tied to one vendor's bundle.
Who it's for
Teams that secure production AI at scale
Burgus Security fits AI platform teams routing high-volume agent traffic, security teams who need OWASP/MITRE-aligned coverage on LLM and MCP connections, compliance owners preparing evidence from live production audit, and engineering leads who want that security without giving up best-of-breed provider choice.
LLM & MCP security
Secure model calls and data connections — check-points, access control, unified audit.
Multi-agent security
Delegation graphs, cross-agent alerts, combined security reports.
Behavioral analysis
Learns normal collaboration automatically on your audit traffic.
Dashboard & MCP
Operate secured production AI — Connect, alerts, usage, programmatic MCP.
Compliance exports
Assessor-ready evidence from the production AI you actually run.
Pricing
USD 189/mo, 50M tokens included, BYOK economics — full pricing details.
Our approach
A layer that watches how agents work together — and signals when they do not
Burgus observes how an organisation's AI agents normally collaborate, learns that baseline, and detects deviation in real time on live API traffic. The goal is behavioral detection across the agent mesh, not a catalogue of known prompt tricks.
1 · Model normal collaboration
Build a reliable picture of how agents delegate, call tools, and exchange data — the interplay between agents, not isolated hops.
2 · Learn without drift
Keep updating the baseline without letting an attacker slowly shift it until malicious behavior looks normal.
3 · Alert with evidence
Surfaces high-signal alerts and evidence — you choose whether to block, throttle, or investigate.
Automatic behavioral learning on audit traffic is available now — enabled per tenant alongside the alert layers below. Alert-only: we surface anomalies; you decide response.
Commercial · production
One subscription secures production AI
Commercial package
$189 USD / month
Burgus charges for the security layer — not for reselling models. 30-day trial with full platform access. Payment method required at signup; first subscription charge when the trial ends unless you cancel before then.
Each billing period includes 50 million tokens on the security layer. Usage above 50M is USD 1 per million tokens.
Included in every subscription
- Full LLM + MCP security, check-points, behavioral learning
- Subscription billed monthly in advance; overage billed in arrears
- Upstream model fees stay with your provider (BYOK)
- 99.9% uptime SLA for paying tenants — see Terms & SLA
Overage examples (Burgus only)
| Tokens / month | Total Burgus |
|---|---|
| 50M (included) | USD 189 |
| 100M | USD 239 |
| 200M | USD 339 |
Example: 200M tokens/month through Burgus → USD 339 security layer (189 + 150 overage). Your model invoices stay with your providers.
Production · included in commercial package
Layered alert system for multi-agent AI workloads
Every tenant receives 4,000+ platform check-points — web-standard and agentic-security rules mapped to OWASP, MITRE ATLAS, and related frameworks — applied automatically on registration. Layer your own approved policies on top. All matching is alert-only: nothing is silently blocked in production; you review, triage, and export evidence via the dashboard or MCP.
Platform alerts · always on
4,000+ check-points
- Credentials, PII, injection, exfil patterns
- Agentic tier: jailbreak, tool abuse, cross-agent handoff
- Applied automatically when you register
Standards reporting
- Coverage matrix: OWASP LLM Top 10, CWE, NIST
- Platform vs customer rule split
- Full standards coverage report on demand
Connect in the dashboard
- 40+ secure provider hostnames
- Register upstream keys on Connect — no curl required
- Copy
base_urland verify traffic on Overview
Your policies · optional layer
Custom content policy alerts
- Upload policy documents via dashboard or API
- Human-approved rules before anything applies
- ISO 42001-aligned governance workflow
Advanced alert reporting
- Rollups, dedup, and rule health at scale
- Per-agent and multi-agent security views
- Export bundles for assessors and incident review
Behavioral analysis · always on
Phase 1 · baseline & graph
- Volume, error, and token spikes vs 7-day baselines
- Workflow fan-out, loop suspects, novel vendors
- Realtime queue ingest + hourly batch re-analysis
Phase 2 · similarity
- Lightweight similarity detection on audit traffic
- Near-duplicate bursts and cross-agent templates
- Semantic drift vs learned centroids
Phase 3 · clustering
- Agent outliers and coordinated burst incidents
- Novel graph topology and hub emergence
- Behavior reports in the dashboard and MCP
Evidence & exports
Audit & compliance exports
- Alert traceability to session and workflow
- NIS2 / ISO-oriented evidence bundles
- Multi-agent + behavior rollups in one API report
Combined security report
- Combined multi-agent security report
- Content policy + behavioral alert summary
- Per-agent anomaly and propagation hints
Production service with transparent pricing.
- Trial: 30 days, payment method at signup — same production infrastructure as paying tenants.
- After trial: USD 189/month including 50M tokens; USD 1/M overage (see terms).
- LLM API usage is additionally billed by your upstream provider (bring your own keys).
- Alert-only: we surface matches; you decide response. SLA applies to paying tenants.
- On-premise and customised deployments available on request — contact us.
FAQ
Common questions about Burgus AI Security & Audit
What is Burgus AI Security & Audit?
A cybersecurity and audit watchpoint for multi-agent AI workloads. The layered alert system combines platform check-points mapped to OWASP and MITRE ATLAS, optional custom policies, always-on behavioral analysis (automatic learning), and exportable evidence for NIS2, ISO 27001, and ISO 42001 procedures.
How much does Burgus Security cost?
New tenants receive a 30-day trial with full platform access. A payment method is collected at signup; the first USD 189/month subscription charge is taken when the trial ends unless you cancel before then. Each billing period includes 50 million tokens through the Burgus security layer; usage above that is USD 1 per million tokens. Upstream model API fees remain with your provider (bring your own keys). See Terms & SLA.
What SLA do you offer?
Paying commercial tenants receive a 99.9% monthly uptime SLA on the security layer data plane and customer API/dashboard, with service credits if we miss the target. Trial tenants use production infrastructure with best-effort support. Full details: Terms & SLA.
How do I connect agents to Burgus?
Sign in at app.burgus.ai with Google or Microsoft,
create an API key, and call the MCP documented at
app.burgus.ai/integrations#mcp.
Point your provider SDK base_url at assigned
{provider}.securellm.burgus.ai secure hostnames.
What is behavioral analysis in Burgus?
Always-on anomaly detection on your audit traffic — not prompt rules. Automatic behavioral learning across three phases: baseline and graph signals, similarity matching, and clustering for coordinated bursts. Enabled at registration; alert-only, with reports in the dashboard and MCP.
Can we use different providers for different agents?
Yes. Burgus is built for multi-vendor production. Register the keys you use in Connect and route each agent through the secure hostname that matches your stack — Groq for speed, Claude for reasoning, or any supported provider.
Do we keep our existing provider keys and contracts?
Yes. Bring your own keys is a consequence of how Burgus secures traffic — independently of any vendor. Model fees stay on your direct invoices; Burgus bills for security only — see pricing.
What counts toward the 50M included tokens?
Tokens processed through Burgus securellm and mcpdata hostnames, counted from
audit metadata (prompt + completion). Your separate OpenAI or Groq invoices are unchanged.
Does Burgus secure MCP tool and data connections?
Yes. The MCP data plane covers tool ingress and egress through *.mcpdata.burgus.ai, with
Access control and full audit — see agent access.
What is included in the USD 189 subscription?
LLM and MCP security, 4,000+ platform check-points, behavioral learning, custom policy workflow, dashboard, MCP, compliance exports, 50M tokens per billing period, and SLA-backed production infrastructure. Full breakdown →
Does Burgus block traffic when alerts fire?
No. Matching is alert-only: Burgus surfaces credentials, injection, agentic abuse, and policy violations for review and export. You decide response; production traffic is not silently blocked.
Contact
Ready to secure your production AI?
Start a 30-day trial at
app.burgus.ai — full platform access, payment method at signup.
Need something different? Contact us directly for on-premise installation, a customised deployment, integration into your environment, or to report an issue. Tell us what you are trying to accomplish and we will respond personally.
Commercial package: 30-day trial, then USD 189/month with 50M tokens included;
USD 1/M overage. See Terms & SLA.
Message sent. We'll be in touch.