OpenAI API and Dome
One OpenAI key serves every agent, and every call is tied to a person. Connect the OpenAI API to Dome once. Each call then names the agent that made it and the verified person it was for, and OpenAI's key never leaves Dome's vault.
Dome provides model brokeringand routing for OpenAI.
OpenAI's key stays in the vault
Dome adds it as a Bearer header when a call goes out, and each read is audited.
Every call made for someone
A rule can refuse any call without a verified person behind it. Audit then names who each call was for.
Spend by agent, not by key
One OpenAI key carries every agent's usage. Dome splits it by agent and person, and a quota stops a loop at its ceiling.
Connect OpenAIin three steps.
Add a connection per model, pool it with Azure for failover, and swap the base URL in your OpenAI SDK code.
Add the connection
The provider id is openai. Dome fills in https://api.openai.com and adds the Bearer prefix to your key.
shelldome models add gpt-openai \ --provider openai \ --model gpt-6-sol \ --api-key "$OPENAI_API_KEY" \ --gateway prod-gatewayPool it with Azure for failover
Add an Azure OpenAI connection for the same model and pool the two, so agents send the pool's name.
shelldome models pool create gpt-6-sol \ --failover-max all --gateway prod-gateway dome models pool member add gpt-6-sol gpt-openai --priority 0dome models pool member add gpt-6-sol gpt-azure --priority 1Point your agent at the Gateway
The OpenAI SDK takes the Gateway URL with /v1 as its base. The agent's Dome key replaces the OpenAI key.
pythonfrom openai import OpenAI client = OpenAI(base_url=f"{GATEWAY_URL}/v1", api_key=DOME_AGENT_KEY)client.responses.create( model="gpt-6-sol", input="Draft a reply to this ticket.",)
No call without a personbehind it.
Applied to an agent, this refuses any OpenAI call that doesn't carry a verified acting-as identity. Unattended runs need their own agent.
cedar
forbid (principal, action == Dome::Action::"llm:invoke", resource is Dome::LLMModel)unless { principal has act_as };Bringing it alltogether.
Connecting OpenAI to registered agents, tools, and identity in Dome completes a governed agent application.
Acting for
Control point
GatewayEvery call decided and audited
RulesGuardsQuotas
Models
OpenAIProvider
GPTModelCommon questionsabout OpenAI.
More of whatDome works with.
Model
Claude Fable
Fable 5.1 from Anthropic and Amazon Bedrock in one failover pool, open to one group and capped by quota.
MCP server
GitHub
The GitHub MCP server behind the Tool Gateway, with rules that decide each call on its owner and repository.
Identity
Okta
Okta tokens verified on every agent call, so rules and audit name the person each agent acted for.
Runtime
LangGraph
LangGraph agents with their model calls on the Model Broker and their MCP tools on the Tool Gateway.
Client
Claude Code
Claude Code on a Dome Gateway with per-developer sign-in, rules on every tool call, and audit by name.
Agent service
TinyFish
TinyFish's web agents behind the Tool Gateway, with rules that decide each run on the site it targets.
Talk with ourFDE team.
Or start free, and register your first agent in minutes. Connect it through a gateway, give it rules, and see every call it makes in one audit trail.
Talk with our FDE team
Our engineers design, build and run governed agents for the part of your business you want to change. You see what they deliver.
LLM providers for AI agents: one governed path to every model
Connect a provider once. Its key stays in Dome, and every agent call to it is authorized, metered and audited.