Docs/Platform

Platform overview

live

FlowRail is the security decision layer for the AI-native SDLC. It embeds security guardrails into spec-driven development (design review through code generation today, scanning and triage on the v1 roadmap, runtime guardrails after that) so AI-generated code is secure by default rather than secure by afterthought.

01

What ships

FlowRail is built as a set of interlocking capabilities. Each has its own page; together they cover the SDLC from intent to incident.

flowrailv0.4.2
1. Design review          read the spec, predict threats, open a review id
2. Guardrail engine       policy framework: OWASP, framework, compliance, custom
3. Pre-write verifier     block risky writes before they land
4. Dependency integrity   gate installs on supply-chain signals
5. Scanner orchestrator   ingest scanner findings, contextually filtered · v1
6. Triage & feedback      auto-triage with guardrail-escape detection · v1
02

What ties them together

In v1, every component writes to the same threat graph, scoped by design_review_id. A finding from your scanner can be traced back to the design-time threat that should have caught it; a guardrail can be tightened automatically when its escape rate climbs. The graph is what makes them more than the sum of their parts. Today, every decision already lands in the append-only audit lineage, threaded by the same design_review_id.

03

Three audiences

Developers: ship faster without security friction. Pre-write feedback in seconds, not post-deployment surprises.

Security teams: in v1, query the threat graph for triage, dashboards, and audit. "Show me all live code derived from threats we flagged in Q1." "What's our guardrail escape rate by category?"

Security leaders: in v1, tune the autonomy slider. Guardrail enforcement (hard block, soft block, advisory). Auto-triage verdicts (autonomous, notify, human-approves). Guardrail updates (confidence-based auto, notify, always-approve).

04

What FlowRail is not

Not a SAST tool (we'll orchestrate Semgrep in v1; we don't replace it). Not a vulnerability scanner (we ingest findings; we don't generate them). Not a code review tool. Scanning is commoditizing; the value is in knowing what to scan, how to triage, and when to act. That's the decision layer we ship.