DeepSeek and Dome
Run DeepSeek V4 Pro, and decide where it runs. With DeepSeek, the question is where it runs. The same weights are on DeepSeek's own API and on US hosts like Together AI and DeepInfra. Tag each connection with its jurisdiction, and a rule keeps regulated work on the ones tagged us.
Dome provides model brokeringand routing for DeepSeek.
One name, two hosts
Together AI and DeepInfra both serve deepseek-ai/DeepSeek-V4-Pro-0813. Pool them as deepseek-v4-pro, and a failed call moves to the next.
Regulated work stays put
Tag every connection with a jurisdiction. A rule on the claims agent refuses any connection not tagged us.
A ceiling on spend
The US pool gets $250 of provider spend a month. The direct DeepSeek connection sits outside it.
DeepSeek with failoverin five steps.
Tag every connection with a jurisdiction, pool the US ones, and scope the claims agent to them.
Connect Together AI and DeepInfra
Both use the same model id. The jurisdiction attribute is your label, and rules read it.
shelldome models add deepseek-together \ --provider together \ --model deepseek-ai/DeepSeek-V4-Pro-0813 \ --api-key "$TOGETHER_API_KEY" \ --attributes '{"family":"deepseek","jurisdiction":"us"}' \ --gateway prod-gateway dome models add deepseek-deepinfra \ --provider deepinfra \ --model deepseek-ai/DeepSeek-V4-Pro-0813 \ --api-key "$DEEPINFRA_API_KEY" \ --attributes '{"family":"deepseek","jurisdiction":"us"}' \ --gateway prod-gatewayConnect DeepSeek directly
For agents with no residency limit. DeepSeek's own id for the model is deepseek-v4-pro.
shelldome models add deepseek-direct \ --provider deepseek \ --model deepseek-v4-pro \ --api-key "$DEEPSEEK_API_KEY" \ --attributes '{"family":"deepseek","jurisdiction":"cn"}' \ --gateway prod-gatewayPool the US connections
Only the us connections join deepseek-v4-pro. Together leads, and DeepInfra takes over when it fails.
shelldome models pool create deepseek-v4-pro \ --failover-max all --gateway prod-gateway dome models pool member add deepseek-v4-pro deepseek-together --priority 0dome models pool member add deepseek-v4-pro deepseek-deepinfra --priority 1Point your agent at the Gateway
The claims agent sends deepseek-v4-pro through the OpenAI SDK, with a Dome key and the Gateway URL.
pythonfrom openai import OpenAI client = OpenAI(base_url=f"{GATEWAY_URL}/v1", api_key=DOME_AGENT_KEY)client.chat.completions.create( model="deepseek-v4-pro", messages=[{"role": "user", "content": "Check this claim against the policy terms."}],)Cap the spend
The quota counts Together and DeepInfra together, as one US budget.
shelldome quotas set --subject pool --pool deepseek-v4-pro \ --unit provider_usd --limit 250 --window monthly
Regulated work stayson connections tagged us.
Applied to an agent with --agent, this refuses any connection not tagged jurisdiction: us. The claims agent can use the pool, never DeepSeek's own API.
cedar
forbid (principal, action == Dome::Action::"llm:invoke", resource is Dome::LLMModel)unless { resource has jurisdiction && resource.jurisdiction == "us" };One call,two outcomes.
Switch the caller or the argument and watch the same call decide differently, with every decision landing in audit.
Model
agent claims-agent · acting as k.osei
llm:invoke(model: "deepseek-v4-pro")
- Agentclaims-agent is registered and active
- Callerk.osei verified
- RuleConnection is tagged jurisdiction: us
DecisionAllowed
Bringing it alltogether.
Connecting DeepSeek to registered agents, tools, and identity in Dome completes a governed agent application.
Acting for
Control point
GatewayEvery call decided and audited
RulesGuardsQuotas
Models
Common questionsabout DeepSeek.
More of whatDome works with.
Provider
Anthropic
The Claude API behind the Model Broker: the key held in Dome, every call authorized, metered and audited.
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.
AI models for enterprise agents: failover, rules and quotas
Every model your agents call goes through the Model Broker. Pool providers for failover, decide who may call each model, and cap what it costs.