What is Agent Engine in Vertex AI?

Vertex AI Agent Engine is Google Cloud's fully managed runtime for deploying, operating and scaling AI agents. It began as LangChain on Vertex AI, took the Agent Engine name on March 4, 2025, and on April 22, 2026 became Google Agent Runtime, part of Gemini Enterprise Agent Platform. You deploy to it from a Google Cloud project, with ready-made paths for ADK, LangChain, LangGraph, AG2 and LlamaIndex agents.

Can I use Gemini models on camelRun?

Yes. With your Google (Gemini API) key saved under Models & keys, an agent can use a model such as google/gemini-3.5-flash, and Google bills you for those tokens directly. Agent time and storage are billed as usual. Gemini models are also in the catalog through OpenRouter.

Can I run AI agents on camelRun outside Google Cloud?

Yes. The TypeScript and Python SDKs reach camelRun over HTTPS from any machine your code runs on, inside Google Cloud or not.

How do I move an ADK agent to camelRun?

Bring over its instruction, its function tools and its model name. The tools become @tool functions in your own app, so they keep running where your app runs, and the agent's key takes the place of the session: the same key is the same agent, with its history, from any process.

What does an AI agent cost to run on camelRun?

camelRun charges $0.01 per agent-hour while an agent is in a run and $0.10 per GB-month of storage, so idle and sleeping agents cost only storage. It's prepaid credit, with no subscription, no seats and no per-agent fee.

Can I move agents between run.camelai.com and my own server?

Yes. Export your account as one zip, then make each agent on your own runtime with the same key, passing its history along. Your app's code stays the same: point the SDK at the new runtime with CAMELAI_BASE_URL.

Compare · camelRun by camelAI

Vertex AI Agent
Engine
alternative

Run AI agents on Gemini or any model, from any cloud

camelRun by camelAI runs the agent loop. Your tools stay in your app, wherever it runs.

camelRun by camelAI runs the agent loop for your AI agents. Your tools stay in your app, on Google Cloud, another cloud or a laptop, and the runtime lives on run.camelai.com or on your own server.

Get an API key Read the quickstart

Vertex AI Agent Engine is now Google Agent Runtime

Your app, wherever it runs

Google Cloud

us-central1

AWS

after the move

Your laptop

localhost

camelRun

Agentasleep

inventory

google/gemini-3.5-flash

run.camelai.com

Tool calls, run in your app

→ read_inventory()

History, kept by camelRun3 of 3 turns

01
Google CloudPlan restocks for everything below target.
inventory Planned BEAN-01 +32 and OAT-02 +24. Cups are above target.
02
AWSWhat did you plan for oat milk?
inventory 24 more, to bring it from 6 up to its target of 30.
03
Your laptopIs anything else below target?
inventory Only the two I planned for: BEAN-01 and OAT-02.

Google's renames

From Vertex AI Agent Engine to Google Agent Runtime

Google Cloud's managed runtime for deploying and scaling AI agents has changed names twice. Guides and search results still use the old ones.

LangChain on Vertex AI
Mar 4, 2025
Renamed, and generally available the same day.
Vertex AI Agent Engine
Apr 22, 2026
Vertex AI becomes part of Gemini Enterprise Agent Platform.
Google Agent Runtime
on Gemini Enterprise Agent Platform

LangChain on Vertex AI
Mar 4, 2025
Renamed, and generally available the same day.
Vertex AI Agent Engine
Apr 22, 2026
Vertex AI becomes part of Gemini Enterprise Agent Platform.
Google Agent Runtime
on Gemini Enterprise Agent Platform

Move an ADK agent to camelRun

From ADK to camelRun

Move an ADK agent to camelRun in Python

On Google's Agent Runtime you deploy your agent's code into a Google Cloud project. On camelRun your tools are plain Python functions that stay in your app, and the runtime runs the loop around them.

Function tools: @tool functions that run in your app, wherever it runs.
Instruction: instructions= on upsert, in plain text.
Model: Any model in the catalog, Gemini included.
Sessions: The agent's key. Same key, same agent and history, with nothing stored on your side.

examples/inventory.py

Python

14from camelai_run import Agents, ToolContext, tool

28    @tool
29    async def read_inventory():
30        """Read inventory and target stock from the application's SQLite database."""
31        return [dict(row) for row in db.execute("SELECT * FROM inventory ORDER BY sku")]

45    async with Agents() as agents:
46        agent = await agents.upsert("inventory", instructions="You manage cafe stock. Plan restocks without purchasing anything.",
47                                    tools=[read_inventory, plan_restock], **({"model": os.environ["AGENT_MODEL"]} if os.environ.get("AGENT_MODEL") else {}))

28    @tool
29    async def read_inventory():
30        """Read inventory and target stock from the application's SQLite database."""
31        return [dict(row) for row in db.execute("SELECT * FROM inventory ORDER BY sku")]
32
33    @tool
34    def plan_restock(sku: str, quantity: int, context: ToolContext):
35        """Set one SKU's local restock quantity. Does not place orders."""
36        if type(quantity) is not int or not 1 <= quantity <= 100:
37            raise ValueError("Invalid restock quantity")
38        if not db.execute("SELECT 1 FROM inventory WHERE sku = ?", (sku,)).fetchone():
39            raise ValueError("Unknown SKU")
40        # The call's stable key: a retried call rewrites the same row.
41        db.execute("INSERT OR REPLACE INTO restock VALUES (?, ?, ?)", (sku, quantity, context.idempotency_key))
42        db.commit()
43        return {"planned": True, "sku": sku, "quantity": quantity}
44
45    async with Agents() as agents:
46        agent = await agents.upsert("inventory", instructions="You manage cafe stock. Plan restocks without purchasing anything.",
47                                    tools=[read_inventory, plan_restock], **({"model": os.environ["AGENT_MODEL"]} if os.environ.get("AGENT_MODEL") else {}))

The whole file on GitHub

Terminal

$ AGENT_MODEL=google/gemini-3.5-flash CAMELAI_API_KEY=art_... python examples/inventory.py --prompt "Plan restocks for everything below target."

Python + SQLite inventory → hosted agent client_…

Planned BEAN-01 +32 and OAT-02 +24. Cups are above target.

{
  "localRestockPlan": [
    {
      "sku": "BEAN-01",
      "quantity": 32
    },
    {
      "sku": "OAT-02",
      "quantity": 24
    }
  ]
}

Models and keys

Use Gemini models on camelRun with your own Google key

Add your Google (Gemini API) key once in the console, and any agent can name a Gemini model. camelRun charges nothing for tokens on your own key.

Switch the same agent to Claude or any other model in the catalog, and its history carries over.
No Google key? Gemini models also run through OpenRouter on our key, charged to your credit.

Read the models and keys guide

The console

run.camelai.com/console/models

Models & keys

Add a key for each provider you want to use. Agents in your account call models with your keys.

Provider keys

Provider	Key	Updated
anthropic	Platform key · billed to credit		Add key
google	…Qk7f	Oct 23, 10:41	Replace
openrouter	Platform key · billed to credit		Add key

Models

Pass the ID as model when creating or configuring an agent.

google

Model ID	Context
google/gemini-3.5-flashGemini 3.5 Flash · reasoning · images	1M	Usable
google/gemini-3.8-flashGemini 3.8 Flash · reasoning · images	1M	Usable

Open source

Self-host camelRun on your own server

camelRun is open source under the AGPL. Self-host it on any server with Docker: one container and Postgres, on whichever cloud you like.

Your app talks to it the same way it talks to run.camelai.com. Point the SDK at it with CAMELAI_BASE_URL and the inventory code above runs unchanged.
Export everything camelRun stores for your account as one zip, from the console, whenever you like.

Self-hosting in the docs

Your server

92export CAMELAI_BASE_URL=http://localhost:8790   # http://runtime:8790 from the Compose network
93export CAMELAI_API_KEY="$AGENT_OPERATOR_TOKEN"
94npx -y @camelai/camelrun whoami                 # the tenant, and the default model

Choosing

Google Agent Runtime or camelRun for your AI agents

Google Agent Runtime fits when

Your agents are built with ADK, LangChain or LangGraph and your team already works in Google Cloud. You deploy from your Google project, with ready-made paths for those frameworks.

camelRun fits when

Your app runs on more than one cloud, or might move. You want Gemini next to Claude and GPT on one agent, an open-source runtime you can host yourself, and prepaid billing with no subscription.

Related camelRun pages

Our guide to AI agent runtimes and where to host them

FAQ

Questions about Vertex AI Agent Engine and camelRun

01
What is Agent Engine in Vertex AI?
Vertex AI Agent Engine is Google Cloud's fully managed runtime for deploying, operating and scaling AI agents. It began as LangChain on Vertex AI, took the Agent Engine name on March 4, 2025, and on April 22, 2026 became Google Agent Runtime, part of Gemini Enterprise Agent Platform. You deploy to it from a Google Cloud project, with ready-made paths for ADK, LangChain, LangGraph, AG2 and LlamaIndex agents.
02
Can I use Gemini models on camelRun?
Yes. With your Google (Gemini API) key saved under Models & keys, an agent can use a model such as google/gemini-3.5-flash, and Google bills you for those tokens directly. Agent time and storage are billed as usual. Gemini models are also in the catalog through OpenRouter.
03
Can I run AI agents on camelRun outside Google Cloud?
Yes. The TypeScript and Python SDKs reach camelRun over HTTPS from any machine your code runs on, inside Google Cloud or not.
04
How do I move an ADK agent to camelRun?
Bring over its instruction, its function tools and its model name. The tools become @tool functions in your own app, so they keep running where your app runs, and the agent's key takes the place of the session: the same key is the same agent, with its history, from any process.
05
What does an AI agent cost to run on camelRun?
camelRun charges $0.01 per agent-hour while an agent is in a run and $0.10 per GB-month of storage, so idle and sleeping agents cost only storage. It's prepaid credit, with no subscription, no seats and no per-agent fee.
06
Can I move agents between run.camelai.com and my own server?
Yes. Export your account as one zip, then make each agent on your own runtime with the same key, passing its history along. Your app's code stays the same: point the SDK at the new runtime with CAMELAI_BASE_URL.

Get started

Run AI agents from any cloud

Get an API key, add your Google key if you want Gemini, and run your first agent from wherever your app lives today.