Claude Sonnet and Dome
Run Claude Sonnet 5.5 from two providers under one name. Sonnet is where many agent teams settle: the code reviewer, the docs writer, and the support agent. The risk is drift, where one config change puts a busy agent on Opus. Dome pins agents to Sonnet and keeps it answering when Anthropic doesn't.
Dome provides model brokeringand routing for Claude Sonnet.
Anthropic first, Bedrock next
Every Sonnet call tries Anthropic at priority 0. Bedrock picks up the ones that fail, under the same model name.
An agent that stays on Sonnet
Tag Sonnet connections with a family attribute. A rule on the agent refuses every model outside it.
$300 a month, all providers
The quota sits on the pool, so it counts Anthropic and Bedrock together. Past $300, Sonnet calls stop until next month.
Sonnet with failoverin five steps.
Tag both Sonnet connections, pool them, pin the code reviewer, then set the monthly cap.
Connect Anthropic
The family: sonnet tag is what the pinning rule checks. The key itself never reaches an agent.
shelldome models add sonnet-55-anthropic \ --provider anthropic \ --model claude-sonnet-5-5 \ --api-key "$ANTHROPIC_API_KEY" \ --attributes '{"family":"sonnet"}' \ --gateway prod-gatewayConnect Amazon Bedrock
Bedrock takes AWS keys, so this one is a dashboard step: region, us.anthropic.claude-sonnet-5-5, and the same family tag. The CLI has no flag for AWS keys.
Pool them for failover
The pool is named for the model, so a new Sonnet is a member swap and no agent needs a release.
shelldome models pool create claude-sonnet-5-5 \ --failover-max all --gateway prod-gateway dome models pool member add claude-sonnet-5-5 sonnet-55-anthropic --priority 0dome models pool member add claude-sonnet-5-5 sonnet-55-bedrock --priority 1Point your agent at the Gateway
The Anthropic SDK works as is. Point it at the Gateway with the code reviewer's Dome key.
pythonimport anthropic client = anthropic.Anthropic(base_url=GATEWAY_URL, api_key=DOME_AGENT_KEY)client.messages.create( model="claude-sonnet-5-5", max_tokens=2048, messages=[{"role": "user", "content": "Review this diff."}],)Cap the spend
A pool quota counts every call through the pool, whichever provider answers.
shelldome quotas set --subject pool --pool claude-sonnet-5-5 \ --unit provider_usd --limit 300 --window monthly
Keep an agenton Sonnet.
Applied to one agent with --agent, this refuses any connection not tagged as Sonnet. The code reviewer can't reach Opus, whatever model name it sends.
cedar
forbid (principal, action == Dome::Action::"llm:invoke", resource is Dome::LLMModel)unless { resource has family && resource.family == "sonnet" };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 code-reviewer · acting as d.chen
llm:invoke(model: "claude-sonnet-5-5")
- Agentcode-reviewer is registered and active
- Callerd.chen verified
- RuleConnection is tagged family: sonnet
- Quota$88 of $300 this month
DecisionAllowed
Bringing it alltogether.
Connecting Claude Sonnet to registered agents, tools, and identity in Dome completes a governed agent application.
Acting for
Agent
Control point
GatewayEvery call decided and audited
RulesGuardsQuotas
Models
Common questionsabout Claude Sonnet.
More of whatDome works with.
Provider
OpenAI
The OpenAI 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
Codex
Codex 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.