Describe
Pick a source platform, a destination, and tell us in plain English what data should move and how often.
Vendor integrations in cybersecurity take 8–12 months. Rinox writes the connector — logging, auth, retries, state, dedup — in under 30 seconds, following a battle-tested seven-section template.
Every integration follows the RINOX IRON standard — fetch, translate, deliver, save state — guaranteeing at-least-once delivery and zero data loss between cron runs.
Pick a source platform, a destination, and tell us in plain English what data should move and how often.
Claude writes the connector against your platform's API docs, following the seven-section RINOX template and the IRON spec.
Download .py / .sh, drop in env vars, schedule it. README ships with cron examples and failure-mode notes.
Beyond source-to-destination connectors, Rinox wraps any single platform as an agent-callable MCP server — typed tools, env-based auth, write-gating, structured errors — in Python or TypeScript, graded against the MCP IRON standard.
Five phases and five appendices, derived from real production failures. Every generation is graded against them before it leaves the API.
Read the full specGenerated scripts follow the seven-section Rinox template and pass the RINOX IRON standard (Phase 0 pipeline plus Appendices A-E). For platforms with uploaded API documentation, accuracy is high enough that most scripts run after only environment-variable configuration. For platforms relying on the model's training knowledge, we render a disclaimer recommending you verify endpoints against your specific version before production use.
API keys, OAuth2 client-credentials, basic auth, bearer tokens, HMAC-signed requests, mutual TLS, and platform-specific schemes such as Splunk HEC tokens, Falcon API client/secret pairs, and Sentinel app-only auth via Entra ID. Credentials are always loaded from environment variables, never hardcoded.
Not yet. Today Rinox runs on rinox.io. Enterprise self-hosting is on the roadmap. If your environment requires it, reach out via the contact form.
Generation prompts and outputs are retained for up to 30 days to support troubleshooting and quality improvement. After that, generation records are removed from production systems. See the security page for the full retention policy.
The active model is configurable from the admin panel and can be switched between Claude Opus, Sonnet, and Haiku without redeploying. The default is Claude Sonnet 4.6. The model in use is recorded with every generation for traceability.
Five free generations to try the product. For team and enterprise pricing — including single-tenant, SSO, and on-prem — get in touch.