Skip to content

API and AI agent security that runs where your data lives.

Cyron detects and blocks attacks on REST, WebSocket, gRPC and streaming APIs from a kernel sensor that adds nothing to your request path. On your own servers, it also protects what your AI agents send to tools and to each other.

Watch a live attack being caught

Free plan, no card · No code changes · EU-hosted SaaS or your own servers

  • 10 of 10 OWASP API Top 10 categories with dedicated detection
  • 31 real-time detectors for streaming and RPC APIs
  • 15 agent-layer detectors for MCP and A2A

Built for teams in

Financial services Healthcare Public sector Critical infrastructure SaaS platforms E-commerce

The attacks that matter look like authorised traffic.

A web firewall checks each request against known patterns. It cannot see a logged-in user reading another customer's records, or an AI agent following an instruction planted in a tool's description. Cyron reads behaviour across the whole conversation.

Over 20%

of organisations reported a breach of their AI models or applications

The joint top cause: a compromised API, application or plug-in, at 27%.

IBM, Cost of a Data Breach Report 2026

25%

of all enterprise generative-AI applications will have at least five minor security incidents a year by 2028, Gartner predicts

Up from 9% in 2025. Gartner ties the rise to agentic applications built on the Model Context Protocol.

Gartner, 9 April 2026

50

sensitive-data patterns in six categories, including national identifiers from 11 countries

Card numbers, IBANs, routing numbers and national IDs are validated to keep alerts clean.

Cyron API Security

Watch three attacks being caught on live traffic.

Three recordings from a live test application: a logged-in user enumerating other customers' orders, a hijacked WebSocket session, and a malicious Protobuf message inside a gRPC stream. Cyron catches each one from a copy of the traffic.

  • APIs monitored: 142 · ↑ 3 added today
  • Threats blocked: 7 · last 24 hrs
  • No added latency · analysis runs on a copy
HTTP Watch on YouTube
Real-time BOLA detection on a live API
Real-time WebSocket threat detection — CSWSH, injection, frame anomalies
Real-time gRPC and protobuf threat detection on wire-level features

Cyron detects an authenticated attacker enumerating order IDs across user accounts; iris then blocks the offending source at the kernel — without slowing legitimate traffic.

One sensor on the host. No change to your code.

  1. Install the iris agent

    Install iris, the eBPF kernel agent, with one command and the token from your dashboard. No SDK, no library to import.

    Attaching eBPF probe to kernel...
✓ Cyron iris agent active

  2. Copy, never intercept

    iris copies your API traffic to the Cyron analysis engine: REST, WebSocket, gRPC and Socket.IO, with Server-Sent Events from launch day. Your live traffic is never held or re-routed. On Essential and above, per-endpoint baselines start forming within hours and sharpen over the first week.

    Copying: REST · WebSocket · gRPC · Socket.IO · SSE
No added latency

  3. Block the attacker at the kernel

    When Cyron confirms an attack, it sends a cryptographically signed block command to iris, which drops the attacker's packets at the kernel. Everyone else's traffic is untouched.

    INCIDENT: BOLA enumeration · API1:2023
Source blocked at kernel

Built for teams that run APIs, and now AI agents.

Growing SaaS teams

Your APIs carry real customer data and you have no security team. Start free, and move up when you want behavioural analysis or the AI analyst.

Free plan, no card. Published prices from $0 to $165 a month.

No security hire needed

Regulated industries

Banks, insurers, hospitals and public bodies whose payloads must stay in-house. Cyron On-Premise runs on your own servers, with the AI analyst on your hardware.

Signed, OCSF-aligned events go to your SIEM. Traffic and findings stay in your network.

On-premise

AI platform teams

You are putting agents into production, with tools over MCP and other agents over A2A. Cyron AI Security protects that traffic and blocks the threat classes you name, with no change to your agents.

MCP and A2A

SOC teams and MSSPs

Every incident reaches your SIEM as a signed, OCSF-aligned event, and borderline events arrive with the AI analyst's verdict and numbered reasoning.

Explainable incidents

Ten OWASP API categories. Five protocols. Fifteen agent-layer detectors.

Cyron has dedicated detection for every category of the OWASP API Security Top 10 (2023), and 31 real-time detectors for WebSocket, gRPC, Server-Sent Events and Socket.IO. Rules run on every plan; behavioural models and the AI analyst join on higher plans.

Business logic abuse

Checkout and purchase velocity, coupon abuse and bots: attacks that use your business logic exactly as it was built.

  • API6:2023

Data theft and exfiltration

Broken object-level authorisation, mass assignment and excessive data exposure. Fifty sensitive-data patterns catch personal and financial data leaving in responses, with card numbers, IBANs and national IDs validated to keep alerts clean.

  • API1:2023
  • API3:2023

Account takeover

Credential stuffing, password spraying, brute force, and JWT attacks such as alg:none and key injection.

  • API2:2023

Streaming and RPC attacks

Cross-site WebSocket hijacking, injection inside messages, floods and Protobuf structure anomalies: 31 real-time detectors, plus deep analysis of data leaving a stream.

  • WebSocket
  • gRPC
  • SSE
  • Socket.IO

Agent-layer attacks

Tool-description poisoning, rug-pulls, secrets in tool arguments, poisoned tool responses and A2A session smuggling: 15 detectors in Cyron AI Security.

  • LLM01:2025
  • ASI04
  • ASI07

Protect your APIs with no change to your code.

No library installs. No SDK changes. No application restarts. Cyron's iris eBPF agent runs alongside your existing stack — your developers never need to touch their code.

  1. Pull the iris container — one command

  2. Agent auto-discovers all active API endpoints

  3. Baselines start forming within hours and sharpen over the first week.

  4. Live protection on — threats blocked at the kernel

Attaching eBPF probe to kernel...
Discovering endpoints...
Found 142 active endpoints across 6 services
Building baselines... continues in the background (7 days)
✓ Cyron iris agent active
Start free

Try it free. Pay when you are ready.

Start free with no card, or try behavioural intelligence and the AI analyst for 14 days. Every plan includes rule-based detection on all five protocols, sensitive-data detection, threat intelligence and API discovery.

A 14-day trial of behavioural intelligence and the AI analyst.

Start 14-day trial

Free

See what Cyron detects.

$0 /mo
$0 /mo, billed annually
  • Everything in the baseline
  • 5 requests analysed per minute
  • No card, no time limit
Start free

Lite

The baseline at working volume.

$15 /mo
$12 /mo, billed annually
  • Everything in Free
  • 60 requests analysed per minute
Start 14-day trial
Most popular

Essential

For APIs that carry sensitive data or money.

$25 /mo
$21 /mo, billed annually
  • Everything in Lite
  • Behavioural intelligence
  • 13 session-level abuse rules
  • 120 requests per minute
Start 14-day trial

Standard

For production, with a reasoned verdict on every borderline event.

$65 /mo
$55 /mo, billed annually
  • Everything in Essential
  • The AI analyst, System 2 Thinking
  • 3,000 AI analyses a month
  • 300 requests per minute
Start 14-day trial

Premium

For larger volumes of traffic.

$165 /mo
$138 /mo, billed annually
  • Everything in Standard
  • 8,000 AI analyses a month
  • 600 requests per minute
Start 14-day trial

Talk to us

Higher volumes and custom terms.

Custom contact us
  • Analysis rates above Premium
  • Custom terms and onboarding
Talk to us

For context: IBM puts the average cost of a data breach at USD 4.99 million in its 2026 report. — IBM, Cost of a Data Breach Report 2026

Need to test before subscribing? Start with the free plan Start free.

Designed to support your work under

  • GDPR EU data protection
  • NIS2 EU network and information security
  • DORA EU digital operational resilience
  • PCI DSS 4.0 Payment card data
  • HIPAA US healthcare data handling

Frequently asked questions

Do I still need a WAF?

Yes, keep it. A WAF blocks requests that match known patterns at the edge. Cyron works on a copy of your traffic and judges behaviour: who is calling, what they read, and how that differs from normal use. Together they cover far more than either alone.

Do my developers need to change any code?

No. iris, the eBPF kernel agent, runs beside your application. There is no SDK and no library to import.

Where is my data processed?

In SaaS, iris sends a copy of your API traffic to the Cyron platform, hosted in Germany. Cyron keeps what it analyses so you can investigate any incident, and removes secrets such as authorisation headers, tokens and card numbers before the AI analyst sees an event. Your live requests never wait for Cyron. Need everything in-house? Cyron On-Premise keeps it all on your own servers.

What is System 2 Thinking?

Cyron's AI analyst. When an event is borderline, it investigates the context, reads the captured exchange and explains its verdict in numbered steps. The verdict is attached to the incident and sent to your SIEM. Real-time detection and blocking never wait for it.

Does Cyron secure AI agents?

Yes. Cyron AI Security protects MCP and A2A traffic on your own servers with 15 detectors and blocks the threat classes you choose. In SaaS, Cyron recognises MCP and A2A traffic and applies its API-layer protection to MCP; full agent-layer analysis comes with Cyron AI Security on-premise.

What is the difference between Cyron API Security and Cyron AI Security?

Cyron API Security judges API calls: who is calling, what they send and how they behave. Cyron AI Security judges what AI agents exchange with tools and with other agents: tool descriptions, arguments, responses and agent messages. An attack on an agent usually looks like a normal, authorised API call; Cyron AI Security reads the instruction behind it.

Can I run Cyron on my own infrastructure?

Yes. Cyron On-Premise installs on your servers from one offline script. The AI analyst runs on your hardware, and your traffic and findings stay in your network. Cyron AI Security is included.

Which protocols does Cyron cover?

REST and HTTP, WebSocket, gRPC, Server-Sent Events and Socket.IO. On-premise, Cyron AI Security adds MCP over Streamable HTTP and A2A over JSON-RPC.

Do I still need a WAF?

Yes, keep it. A WAF blocks requests that match known patterns at the edge. Cyron works on a copy of your traffic and judges behaviour: who is calling, what they read, and how that differs from normal use. Together they cover far more than either alone.

Do my developers need to change any code?

No. iris, the eBPF kernel agent, runs beside your application. There is no SDK and no library to import.

Where is my data processed?

In SaaS, iris sends a copy of your API traffic to the Cyron platform, hosted in Germany. Cyron keeps what it analyses so you can investigate any incident, and removes secrets such as authorisation headers, tokens and card numbers before the AI analyst sees an event. Your live requests never wait for Cyron. Need everything in-house? Cyron On-Premise keeps it all on your own servers.

What is System 2 Thinking?

Cyron's AI analyst. When an event is borderline, it investigates the context, reads the captured exchange and explains its verdict in numbered steps. The verdict is attached to the incident and sent to your SIEM. Real-time detection and blocking never wait for it.

Does Cyron secure AI agents?

Yes. Cyron AI Security protects MCP and A2A traffic on your own servers with 15 detectors and blocks the threat classes you choose. In SaaS, Cyron recognises MCP and A2A traffic and applies its API-layer protection to MCP; full agent-layer analysis comes with Cyron AI Security on-premise.

What is the difference between Cyron API Security and Cyron AI Security?

Cyron API Security judges API calls: who is calling, what they send and how they behave. Cyron AI Security judges what AI agents exchange with tools and with other agents: tool descriptions, arguments, responses and agent messages. An attack on an agent usually looks like a normal, authorised API call; Cyron AI Security reads the instruction behind it.

Can I run Cyron on my own infrastructure?

Yes. Cyron On-Premise installs on your servers from one offline script. The AI analyst runs on your hardware, and your traffic and findings stay in your network. Cyron AI Security is included.

Which protocols does Cyron cover?

REST and HTTP, WebSocket, gRPC, Server-Sent Events and Socket.IO. On-premise, Cyron AI Security adds MCP over Streamable HTTP and A2A over JSON-RPC.

Your APIs are running right now. Is anything watching them?

Start free: no card, no sales call, no code changes. Need it inside your own network? Ask for an on-premise evaluation.

Start free

Free plan, no card · 14-day trial · EU-hosted SaaS · On-premise for regulated teams