Guide · Agent skill

How to use the Gemini Managed Agents API agent skill

Updated 2026-10-29 · written with AI from the project's own documentation and checked against our catalogue data

This guide explains the gemini-agents-api agent skill, which teaches an AI assistant how to manage custom Agent resources on Google's Gemini Enterprise Agent Platform through its REST API. It is written for developers and platform teams who have not used agent skills before. By the end you will know how to install the skill, ask for work in natural language, and review the requests it produces before you run them.

Gemini Managed Agents APIgoogle/skills · Apache-2.0 Download the skill

Quick start

1Download and unzip. Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or your project's .claude/skills/ folder.
2Check prerequisites. Confirm gcloud is installed and authenticated, and that your Google Cloud project has access to the Managed Agents API.
3Set variables. Export PROJECT_ID, LOCATION set to global, and ACCESS_TOKEN from gcloud auth print-access-token.
4Ask for an agent task. Describe the agent you want in plain language, including the actions and any files, skills or tools.
5Review the payload. Check the tools, network allowlist and environment sources before sending any request.
6Poll the operation. Use the operation name returned by create to check the GET endpoint until done is true.

What is an agent skill and what does this one do?

An agent skill is a folder of instructions and reference material that an AI assistant loads when a task matches its description. This skill covers the control plane of the Managed Agents API: creating, retrieving, listing, updating and deleting stateful, server-managed agent resources before you run conversations with them.

Once the skill is loaded, the assistant can write the exact REST requests for creating an agent with system instructions and tools, mounting files or skills from Google Cloud Storage, and adding external MCP tool servers. It also knows how to poll the long-running operation that creation returns until the agent is ready. It does not chat with Gemini for you; it is a reference for a developer-facing API.

What you need before you install

The skill assumes a working Google Cloud setup. The source lists these requirements: the gcloud CLI installed and authenticated, a Google Cloud project ID with access to the Managed Agents API, and curl or an equivalent HTTP client. Requests use an access token from gcloud and Application Default Credentials.

gcloud CLI installed and authenticated (for gcloud auth print-access-token and Application Default Credentials)
A Google Cloud project ID with access to the Gemini Enterprise Agent Platform Managed Agents API
curl or an equivalent HTTP client
Optional: a Cloud Storage bucket or Skill Registry resource, and any third-party MCP server credentials

Install the skill (ZIP download or Claude apps upload)

Download the ZIP from the skill page on this directory and unzip it. For a personal install, place the folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder. Alternatively, upload the skill in the Claude apps under Settings → Capabilities → Skills.

The package contains SKILL.md, which holds the instructions, and references/attribution.md, which the skill must read before running commands.

Set up your project variables

Before any request, the skill expects a project ID, a location and an access token. The source sets the location to global for the production endpoint and gets the token from gcloud. Replace the placeholder project ID with your own.

export PROJECT_ID="your-project-id"
export LOCATION="global"
export ACCESS_TOKEN=$(gcloud auth print-access-token)

How to trigger the skill with natural requests

You do not need to name the skill. Ask for a concrete agent-resource task in plain language, and mention the project and what the agent should contain. Good requests name an action (create, configure, list, update, delete) and a resource, for example asking for an agent with a system instruction and Cloud Storage files mounted into its workspace.

Be specific about the outcome you want. If you only ask a general question about Gemini, the skill is not the right fit.

Create a custom agent named my-custom-agent in my project with a domain-expert system instruction.
Mount the skills folder from gs://your-agent-bucket-name into the agent's /.agent/skills directory.
Show me how to poll the operation returned when I create an agent.
List and retrieve the configuration of my existing custom agents.

Review the generated request before running it

The source's example create request enables several tools and sets the network allowlist domain to '*', which allows all domains, and it includes code execution and filesystem tools. Those are broad permissions. Read every field of a generated payload, change the allowlist to the domains you actually need, and remove tools the agent does not need.

Creating an agent returns an operation object rather than the finished agent. Use the operation name to poll the GET endpoint until the response shows done set to true, then read the agent details from the response field.

Replace the wildcard network allowlist with specific domains where possible.
Only enable tools such as code_execution and filesystem when the agent needs them.
Keep MCP tokens out of shared files; the skill sends custom headers only to the MCP server URL you specify.

What people use it for

Provisioning a standard set of agents for a team, each with its own system instruction.
Mounting skill files from a Cloud Storage bucket into an agent's workspace.
Adding a third-party MCP tool server to an agent configuration with its own auth header.
Listing, updating or deleting agents as part of an automated deployment pipeline.

Frequently asked questions

Is the Gemini Managed Agents API agent skill free?

The skill is published under the Apache-2.0 licence. The Google Cloud resources it manages are not covered by that licence, so check your own Google Cloud project and billing setup before creating agents.

What do I need to install and use this skill?

You need the gcloud CLI installed and authenticated, a Google Cloud project ID with access to the Managed Agents API, and curl or another HTTP client. Cloud Storage or Skill Registry resources and MCP credentials are optional.

Where do I put the skill folder after downloading?

Unzip it into ~/.claude/skills/ for a personal install or into a project's .claude/skills/ folder for one project. You can also upload it in the Claude apps under Settings → Capabilities → Skills.

Does this skill run scripts or access the network?

The source lists that it does not run scripts but does use the network, since it sends REST requests to the Google API endpoint and any MCP server you configure.

Why does creating an agent not return the agent right away?

Creation is a long-running operation. The endpoint returns an operation object, and you poll that operation until done is true, at which point the agent resource appears in the response field.

Can I use this skill with other AI tools or only Claude?

The source only describes installing it in ~/.claude/skills/, a project's .claude/skills/ folder, or the Claude apps. It does not state support for other tools, so check before assuming it works elsewhere.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use this skill if you are a developer or platform engineer who wants an assistant to write and explain Managed Agents API calls for stateful agents on Google Cloud. Pick something else if you only want to chat with Gemini, or if you cannot set up a Google Cloud project with billing, since every request depends on one.

Download the skill