SecurityMIT
SecGate
One command. One report. One exit code.
Runs Semgrep, Gitleaks, osv-scanner, Trivy, and npm audit in one command. Normalizes findings into one report and fails the pipeline on CRITICAL or HIGH.
Security infrastructure built for
engineers who ship.
Stelnyx gives you scanning, memory, and agent tooling that doesn't slow your team down.
Tools
Everything we build, shipped publicly.
Infrastructure tools built for engineers. MIT where open, proprietary where not.
Codebase Intel
Lumen
Codebase intelligence for engineers.
AST analysis, risk scoring, architecture diagrams, docs generation, and MCP support. Local-first — bring your own AI key. No cloud required.
Agent Readiness
LuxFaber
Lighthouse for the agent web.
Scans any URL and returns a scored report on how readable and usable the page is for LLM-driven agents. Static HTTP fetch — no JS rendering, no telemetry, no account.
Issue ForgeMIT
Intake
Paste anything. Get a structured issue.
Paste a Slack thread, bug report, or vague complaint — Intake uses local AI via Ollama to return a structured GitHub issue in one command. No cloud, no account.
Memory LayerMIT
Engram
One ledger. One confidence model.
Memory ledger for AI systems. Sessions, facts, confidence scoring, and MCP support. Local-first — one context budget, no account, no telemetry.
Agent GuardMIT
Arbiter
One guard. One verdict. Sub-5ms.
Agent guard layer — enforces limits, validates schemas, checks tool allowlists, tracks cost, and detects loops. No LLM required. Deterministic and fast.
Philosophy
The name Stelnyx comes from two directions: stellar light and the Greek goddess of night. Every tool we build has to work in the dark, when systems fail at 2am, and also in daylight, when you're making deliberate choices about architecture. We build at the infrastructure layer because that's where leverage is.