Glimrel team@glimrel.store Get a gateway key

About

A Bombay workshop for model traffic.

Glimrel is a gateway, trace, and eval plane for LLM traffic. Point your SDK at one endpoint to route models, watch spend and latency, and score production spans against the same graders you use in CI.

Founder

Ujjwal Kumar Singh builds Glimrel full time. He studied Computer Science and Engineering at Indian Institute of Technology Bombay.

Security

TLS everywhere, SSO on Flood and above, logs retained 90 days by default. Ask for the latest SOC packet by email — we do not publish a certification claim on this page.

AI posture

A control plane, not an agent company.

Glimrel’s industry focus is AI development tools / APIs. The core product is multi-model routing, nested LLM traces, and LLM-as-judge evals. We are not building AI agents. Planned model access is Bedrock, Anthropic, OpenAI direct, Groq, and Hugging Face.

What models do here

Providers generate completions. Glimrel routes them, traces the tree, and can run an LLM judge on a sample. We do not train Claude, GPT-4o, Gemini, Llama, Nova, or Groq.

What operators do

Set fallbacks, spend caps, and graders. Production pins that config. Agents you already ship — LangChain, Assistants, homebrew — point at one key.

Stage

Initial product / MVP. Uptime targets live on the order form. Indian law governs, with courts in the campus city of the founder.

Privacy

Glimrel stores account email, project config, and the payloads you send to the API. We do not sell customer data. Delete requests complete within 30 days of a note to team@glimrel.store.

Terms

The hosted service is provided as an MVP. Uptime targets live on the order form. Indian law governs, with courts in the campus city of the founder.

Mail

Write the team

Notes land with Ujjwal at team@glimrel.store. Use the form if you prefer to leave a draft on this page first.