AI products you can use today

LLM gateway · Windows desktop

Stop overpaying for intelligence.

A local, OpenAI-compatible API router at localhost:4242. It scores each prompt’s complexity from 0 to 100 and sends it to the cheapest model that qualifies — local Ollama or LM Studio first, OpenAI, Anthropic or Groq only when needed.

• Seven routing modes: cost, privacy, coding and more
• Drop-in for Claude Code, Cursor and Continue.dev
• Shows why each prompt went where it did
• Prompt and response content is never stored

Enterprise LLM gateway · Türkiye

One endpoint. Every approved model. Full control.

A multi-tenant, OpenAI-compatible LLM gateway for organizations in Türkiye. Your apps call one endpoint; RotaLLM decides which Turkish-hosted providers and models each team may use, how much they can spend and what content gets through.

• Model policies at organization, group, member and key level
• Budgets, rate limits and usage metering with CSV export
• Turkish-aware PII guardrails: TCKN, IBAN, phone, email
• Tenant isolation enforced in the database (row-level security)
• Bilingual TR/EN dashboard for admins and members

AI agent · News alerts

Tell it what to watch for.

Describe an event in one sentence. CueBot works out where to look, checks every source before using it, and alerts you only when the event actually happens — not when it is merely mentioned.

• Explains why each alert matched
• Tells you when a watch finds nothing, so you know it’s working
• Email, Telegram and WhatsApp delivery

News aggregator · AI/ML

Top GenAI/ML news, curated hourly.

AI agents read the sources every hour and surface what matters, grouped by model releases, research, policy and regulation, industry, and tooling and infrastructure.

• Importance score and HOT label on each story
• Browse any day of the past week, or search a topic
• Source shown on every item

Open source & labs

On GitHub

nazar · Python · pre-alpha Nazar AI-assisted protection for Linux servers, built on CrowdSec. A local model reviews alerts, explains them in plain language and applies bounded, reversible actions — no GPU or cloud LLM needed. single-gpu-llm-bench · Python single-gpu-llm-bench Reproducible local LLM benchmarks on one consumer GPU — speed, context ceiling, GSM8K, MMLU-Pro, HumanEval and GAIA. route42app · Go Route42 gateway The open gateway behind Route42: local-first routing to the cheapest qualified model.