Documentation

Build on every model

SevenLLM is one workspace and one API in front of GPT-5, Claude, Gemini, Llama, Grok, DeepSeek and Mistral. These docs cover the core concepts and how to get going.

Overview

Most teams wire up several model providers, each with its own SDK, billing, and quirks. SevenLLM replaces that with a single workspace and a single endpoint. You pick a task, and we route it to the best model for your plan, with automatic fallback if a provider degrades.

Quickstart

Three steps to your first result:

  1. 1Create a free workspace and sign in (GitHub or an email magic link).
  2. 2Upload a file or paste a prompt, then choose a task: summarize, compare, extract, translate, or Q&A.
  3. 3Run it. Watch live status, then download or share the result.

Models & routing

Seven model families are available today, with new ones added as they ship. The router selects a model by task and plan tier and falls back automatically on provider errors.

GPT-5ClaudeGeminiLlamaGrokDeepSeekMistral

Compare

Run one prompt across several models at once and read the answers side by side, with quality, latency, and cost surfaced together so you can pick the right model in minutes.

Workflows

Chain tasks into a repeatable pipeline (for example: summarize, then extract, then translate). Each step passes its result to the next, runs in the background, and retries independently if a step fails.

API

Every model is reachable through one REST endpoint authenticated with an API key you generate in Settings. Keys are shown once at creation and stored only as a hash. Full request and SDK references are published here as the public API stabilises.

Security & privacy

Data is isolated per tenant with row-level security, encrypted in transit and at rest, and never used to train models. See the Privacy Policy for data handling, sub-processors, and your rights under GDPR and India's DPDP Act.

Start building