Skip to main content
Gloo AI is a platform for building with AI that fits your context. Many people know it for answers grounded in your own content and values-aligned answers with tradition settings. You also get models from OpenAI, Anthropic, Google, and open source providers behind one API, a web app to manage it all, and a coding agent. It is three products on one account and one bill, managed in Studio, so you can start with one and add the others later.

Make your first request

Create an account, fund your wallet with as little as $1, turn on auto-refill, create an API key, and send a request.

​
Three products, one account

Inference and APIs

One OpenAI-compatible API to more than 160 models. Call them directly at the model’s rate, or through Gloo’s guardrails and tradition settings. Includes embeddings, tools, streaming, and provider failover. Pay as you go with prepaid credits.

Data Engine

Bring your own content from files, websites, YouTube channels, and feeds. Search it, and ground answers in it with citations. A plan; currently at capacity for new organizations.

Gloo Code

A coding agent that plans, writes, and tests changes. Use it from your terminal, the desktop app, or headless in scripts. Billed per seat on a monthly plan.

Studio

The web app where you manage all three: try models in the Playground, create API keys, see usage and billing, and manage your organization and team.
New to AI? Start with the Learning Center.

​
What do you want to do?

​
How Gloo Works

Gloo sits between your application and AI model providers, wrapping leading models in a harness for values alignment and guardrails. Call them directly with nothing added, build with them through Gloo’s guardrails, or ingest your trusted content and ground answers in it, with sources cited. Your application talks to one API, which you can use in three ways. Not sure which fits? See Endpoint types.

​
Use the models directly (direct)

The Direct Responses, Direct Completions and Anthropic Messages APIs send your request straight to the model you name. Gloo adds no guardrails, output moderation, or values-alignment, so your application owns its own safety. Authentication, metering, and the Gloo model catalog still apply. Direct endpoints bill the model’s published rate.

​
Use the models with Gloo’s guardrails (guarded)

Your application sends a question to the Gloo AI API, which runs it through third-party AI models wrapped in the Gloo Harness and returns a values-aligned answer Your application sends a question to the Gloo AI API, which runs it through third-party AI models wrapped in the Gloo Harness and returns a values-aligned answer The guarded Responses (and Completions) APIs give you leading third-party models wrapped in Gloo’s harness: the values-alignment, guardrails, and output moderation that make them Gloo’s model lineup rather than raw provider endpoints. Choose guarded when you want Gloo’s safety and values alignment built in. On Completions you also get model families and auto-routing, and Grounded Completions builds on Guarded Completions.

​
Ground answers in your content (grounded)

Grounded flow: your application sends a question to the Gloo AI API, which searches the Data Engine for matching snippets of your trusted content (step 1), then runs inference through third-party AI models in the Gloo Harness with the prompt plus the retrieved snippets (step 2), and returns the answer Grounded flow: your application sends a question to the Gloo AI API, which searches the Data Engine for matching snippets of your trusted content (step 1), then runs inference through third-party AI models in the Gloo Harness with the prompt plus the retrieved snippets (step 2), and returns the answer The Data Engine ingests your content and keeps it current, from files, websites, YouTube channels, and feeds. When you call Grounded Responses (or Grounded Completions), Gloo handles the two-step flow shown above: semantic search finds the snippets of your content most relevant to the question, and the model answers using them, with citations back to your sources. Prefer to manage retrieval yourself? The Search API returns the matching snippets directly, so you can format the context and prompt the models on your own terms. If you manage your own vector store and retrieval pipeline, use the Embeddings API to create vectors with Gloo embedding models.
See the full production journey, from ingesting your content to serving grounded answers and keeping it current, in the End-to-End RAG Pipeline series.

​
What you get with Gloo

  • Direct and guarded endpoints. Call the models directly with nothing added, or use them with Gloo’s guardrails, output moderation, and tradition settings. See Endpoint types.
  • Clear pricing. Direct endpoints bill the model’s published rate (see the Models page and GET /platform/v2/models). Guarded endpoints are priced at that rate plus 6.5% for guardrails, moderation, and values alignment.
  • Grounded answers from your own content. With Data Engine access, Grounded Completions can cite the sources they used.
  • Provider failover. If the provider serving a model is failing, Gloo sends the request to another provider that serves the same model. See Provider failover.
  • A model catalog. More than 160 models across the major provider families, including embedding models, listed live on Supported models.
  • Gloo Code. A coding agent for your terminal on the same account. See Gloo Code.
  • Organizations and roles. Admin, Editor, and Viewer roles control who can manage content and billing. See Manage users.

​
Next steps

Pick the starting point that fits your role.

Developer

Build applications with the Gloo AI APIs and SDKs.

Admin

Set up your organization, publishers, and team.

Publisher & Editor

Upload and manage the content that powers your AI.

Just exploring

New to AI or to Gloo? Start with the Learning Center.