Model router for agentic systems. Routes every prompt to the right model in <50ms. Cut costs 40-70% with just an endpoint change.
Digging Deeper....
A high-speed tool for passively gathering URLs, optimized for efficient and comprehensive web asset discovery without active scanning.
Quick and Easy server testing/validation
Read-only developer endpoint scanner for on-disk package, extension, and developer-tool metadata, built to check exposure to known software supply-chain compromises.
Take a list of domains, crawl urls and scan for endpoints, secrets, api keys, file extensions, tokens and more
Security sensor for realtime threat detection and protection
Visibility into AI agent activity on endpoints, with on-device detection, optional pre-action blocking, and forensic reconstruction.
Self-hosted AI gateway for coding CLIs β one OpenAI/Claude/Gemini/Codex-compatible endpoint, with a multi-tenant web console, request logs, and spend quotas.
Intelligent API gateway for Claude Code and Codex CLI - rotate endpoints, monitor usage, and seamlessly integrate OpenAI, Gemini, and other platforms.
A tool for auditing endpoints defined in exposed (Swagger/OpenAPI) definition files.
Sidecar container that watches Kubernetes Snapshot CRD objects and triggers CreateSnapshot/DeleteSnapshot against a CSI endpoint.
See what every AI coding agent on your machine is doing, and whether the API behind it is the one you paid for. Claude Code, Codex, Cursor, Gemini CLI and more: one endpoint for 44 model vendors, AβF API authenticity grading, per-agent permissions, usage and prompt records.
jshunter is a command-line tool designed for analyzing JavaScript files and extracting endpoints. This tool specializes in identifying sensitive data, such as API endpoints and potential security vulnerabilities, making it an essential resource for and bug bounty hunters and security researchers.
Fast and efficient osquery management
Open-source infrastructure for the full inference lifecycle: deploy, observe, scale, optimize, and safely release self-hosted models behind one endpoint.
Mullvad-first, WireGuard-general: one internal endpoint, hundreds of rotating exit IPs
One local endpoint. The right model for every turn β enforced in code, not prompts.