Akto for AppSec engineers

The scanner operates as a self-service black-box platform. You submit an API endpoint URL and receive a risk score on an A–F scale along with prioritized findings. The scan uses only read-only methods such as GET and HEAD, with text-only POST for LLM probes, and completes in under one minute. No agents, SDKs, or code access are required, and the approach is compatible with any language, framework, or cloud environment.

The tool detects issues across 12 categories aligned to the OWASP API Top 10 (2023), including Authentication bypasses, JWT misconfigurations such as alg=none or expired tokens, BOLA and IDOR via sequential ID enumeration, BFLA and privilege escalation probes, and sensitive data exposure like PII, API keys, and error leakage. It also identifies input validation risks, rate-limiting weaknesses, encryption misconfigurations, SSRF indicators, inventory issues, unsafe consumption surfaces, and LLM/AI security adversarial probes. Findings map directly to PCI-DSS 4.0, SOC 2 Type II, and OWASP API Top 10 (2023).

It parses OpenAPI 3.0, 3.1, and Swagger 2.0 files with recursive $ref resolution, cross-referencing spec definitions against runtime behavior to find undefined security schemes, deprecated operations, and missing pagination. Authenticated scanning supports Bearer, API key, Basic auth, and Cookie methods. Domain verification is enforced through DNS TXT records or an HTTP well-known file, ensuring only the domain owner can run scans with credentials. A strict header allowlist limits forwarded headers to Authorization, X-API-Key, Cookie, and X-Custom-*.

Pro tier features scheduled rescans every 6 hours, daily, weekly, or monthly, with diff detection to surface new findings, resolved items, and score drift. Alerts are rate-limited to one email per hour per API, and HMAC-SHA256 signed webhooks disable automatically after five consecutive failures. The platform integrates via a web dashboard for reports and trend tracking, a CLI with JSON or text output, a GitHub Action that fails builds below a score threshold, an MCP server for AI coding assistants, and a programmatic API for custom workflows.

Scanning is read-only, with destructive payloads never sent. Private IPs, localhost, and cloud metadata endpoints are blocked at multiple layers. Customer data is deletable on demand and purged within 30 days of cancellation; it is never sold or used for model training. The tool does not fix, patch, block, or remediate, nor does it perform active SQL or command injection testing. It does not detect business logic vulnerabilities, blind SSRF, or replace a human pentester for high-stakes audits. It provides findings and remediation guidance only.