Google ADK agents and Dome
Keep your ADK agents, and govern every call they make. ADK is built around Gemini. Through Dome, the same agent can think with Claude or GPT. Swap its model for a LiteLLM model pointed at the Broker and its tools for an MCP toolset on the Gateway. Sub-agents and the runner don't change, and a rule decides every call.
Dome governs every tool and model callin Google ADK.
Sub-agents stay as they are
Agents, instructions, and the runner are unchanged, and only the model and the toolset change.
Any model, not only Gemini
ADK's LiteLLM wrapper talks to the Broker's OpenAI-compatible endpoint. The model name is a pool, so Claude, GPT and Gemini sit behind one setting.
Gemini and Claude in one trail
Whichever pool answers, the model call is recorded beside the agent's tool calls, against the person it acted for.
Google ADK on Domein two changes.
Two edits to the agent's constructor: a LiteLlm model with the Broker as api_base, and an McpToolset on the Gateway.
Install the packages
Install ADK with its MCP extra, and LiteLLM for non-Gemini endpoints, on Python 3.10 or later.
shellpip install "google-adk[mcp]" litellmLoad tools from the Gateway
McpToolset connects over streamable HTTP and lists the tools this agent is granted, schemas included.
pythonfrom google.adk.tools.mcp_tool import McpToolset, StreamableHTTPConnectionParams dome = McpToolset(connection_params=StreamableHTTPConnectionParams( url=f"{GATEWAY_URL}/mcp", headers={"Authorization": f"Bearer {DOME_AGENT_KEY}"},))Call models through the Broker
Use LiteLLM's OpenAI route, set the Gateway as its base, and name a pool as the model.
pythonfrom google.adk.models.lite_llm import LiteLlm model = LiteLlm( model="openai/claude-opus-5-5", api_base=f"{GATEWAY_URL}/v1", api_key=DOME_AGENT_KEY,)
One agent,end to end.
An ADK LlmAgent that runs Claude Opus through LiteLlm and takes its tools from a McpToolset on the Gateway. No Google or Anthropic key in the code.
researcher
Answer questions about the code
InMemoryRunner drives the agent through the repositories. The Gateway decides each tool call, and the Broker sends each model call to a provider that answers.
python
import asyncio
from google.adk.agents import LlmAgentfrom google.adk.models.lite_llm import LiteLlmfrom google.adk.runners import InMemoryRunnerfrom google.adk.tools.mcp_tool import McpToolset, StreamableHTTPConnectionParamsfrom google.genai import types
researcher = LlmAgent( name="researcher", instruction="Answer questions about the code. Search before you answer.", # Model: a pool on the Broker, through the OpenAI-compatible endpoint model=LiteLlm( model="openai/claude-opus-5-5", api_base=f"{GATEWAY_URL}/v1", api_key=DOME_AGENT_KEY, ), # Tools: whatever this agent is granted on the Gateway tools=[McpToolset(connection_params=StreamableHTTPConnectionParams( url=f"{GATEWAY_URL}/mcp", headers={"Authorization": f"Bearer {DOME_AGENT_KEY}"}, ))],)
async def main(): runner = InMemoryRunner(agent=researcher, app_name="researcher") session = await runner.session_service.create_session(app_name="researcher", user_id="k.brandt") question = types.Content(role="user", parts=[types.Part(text="Where is rate limiting handled?")]) async for event in runner.run_async(user_id="k.brandt", session_id=session.id, new_message=question): if event.is_final_response() and event.content: print(event.content.parts[0].text)
asyncio.run(main())Bringing it alltogether.
Connecting Google ADK agents to tools, models, and identity in Dome completes a governed agent application.
Acting for
Agent
Google ADKRuntime
Control point
GatewayEvery call decided and audited
RulesGuardsQuotas
Tools
Models
Common questionsabout Google ADK.
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.
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.
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.
Agent frameworks and runtimes: govern agents without a rewrite
Build agents on the framework you already use. Dome governs their tool and model calls without a rewrite.