GMI AgentBox is where teams find ready-to-use AI agents and where builders publish their own.Documentation Index
Fetch the complete documentation index at: https://docs.gmicloud.ai/llms.txt
Use this file to discover all available pages before exploring further.
Using an agent
- Search the catalog or filter by category.

- Open an agent and read its description and infrastructure badges.

- Click Access agent, or Request Early Access if it’s in beta.
- Sign in with your GMI account. Usage is billed against your account.
Register an agent
- Go to Register an agent in the menu. Pick GMI GPU Clusters if you want GMI to host your agent, or GMI MaaS if you prefer to self-host and use GMI for model inference.

- Open the Deploy Wizard from your GMI dashboard. If you choose GPU Clusters, it builds your container, runs it, and automatically assigns a public URL.
List an agent
- Fill in your agent details.

- Submit for review.
Trust and badges
The badge on an agent card signals who runs the infrastructure and how the listing was reviewed.| Verified | Powered by GMI MaaS | Powered by GMI GPU Clusters |
|---|---|---|
| Reviewed by GMI and running end-to-end on GMI infrastructure. | Self-hosted by the publisher, calling GMI Models-as-a-Service for inference. | Hosted on GMI GPU Clusters on dedicated infrastructure. The publisher supplies the model and API. |



What is an agent?
An agent is a focused AI worker that does one job well. One agent reviews code. Another summarizes PDFs, drafts marketing copy, or answers questions about your support docs. Each agent has a dedicated page on the Marketplace describing what it does, who built it, and how to start using it. Behind every agent sits a model and the application code that turns it into something useful. Both run on GMI Cloud, so you skip the work of provisioning servers, GPUs, and scaling rules.Who uses the Marketplace
| Users | Builders |
|---|---|
| Browse the catalog, find an agent that fits the task, and start using it. | Bring an agent, deploy it on GMI, and list it for others to find and use. |
Pricing
Users pay for what they consume. Model calls bill per token through MaaS. Compute bills per second of agent runtime on GPU Clusters. Builders pay only for the compute their agent uses. There is no listing fee and no minimum spend.Where to start
- Just browsing? Open the catalog and pick any Verified agent.
- Looking for something specific? Use search with a keyword like “code review” or “OCR”.
- Building? Open the Deploy Wizard from your GMI dashboard and start with the managed path.