Model providers
Available now
Halo and Amazon Bedrock
Halo-managed models run through Amazon Bedrock, so agents can use governed models with no provider key.
How Halo works with Amazon Bedrock
Halo-managed models run through Amazon Bedrock in Halo’s managed model path. Model requests pass through the Halo model gateway and its configured governance controls.
Managed usage is billed through Halo credits, so teams do not create, store, or rotate a provider key. To use your own provider account instead, Halo supports bring your own key for OpenAI, Anthropic, Google Gemini, Cerebras, and GMI Cloud.
Governance
What Halo governs for Amazon Bedrock
Managed model path
Route Halo-managed model requests through the model gateway using Amazon Bedrock.
No provider key to manage
Use Halo-managed models without creating, storing, or rotating a provider credential.
Gateway policy
Apply configured checks and approval gates to supported model and agent traffic.
Budget and evidence
Set account or agent budgets and retain reviewable records of decisions and costs.
Example agent jobs
Put agents to work with Amazon Bedrock
Illustrative jobs. Each one runs with scoped access, policy, and evidence.
Illustrative internal summary
For example, an internal agent could summarize an approved request using a model available through the Halo-managed Bedrock path.
Illustrative workflow draft
For example, an agent could draft a response using a Halo-managed model, with policy checks and review for higher-impact actions.
Setup
Get connected in three steps
- 1
Choose an access mode
Select Halo-managed model access when you configure an agent.
- 2
Configure the model
Choose an approved model from the Halo-managed catalog.
- 3
Review the controls
Set policy, approvals, and budgets, then test the agent workflow through the gateway.
FAQ
Questions, answered
Halo-managed models run through Amazon Bedrock in the Halo-managed model path.
Not today. Bedrock powers Halo-managed models, which use Halo credits instead of your own AWS credentials. For your own keys, Halo supports OpenAI, Anthropic, Google Gemini, Cerebras, and GMI Cloud.
Halo-managed model usage is billed through Halo credits, and the account and per-agent budgets you set still apply at the gateway.
Related capabilities
More model providers
Put a halo around the AI your business runs.
See discovery, runtime policy, approvals, and evidence working together on your own use cases.