> For the complete documentation index, see [llms.txt](https://opengpu-network.gitbook.io/opengpu-network/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://opengpu-network.gitbook.io/opengpu-network/introduction/ai-agents.md).

# AI Agents

The Domain of AI Agents represents the next evolutionary layer of the OpenGPU ecosystem—a framework where autonomous AI agents operate as independent economic entities within the network.

## What Are AI Agents?

AI agents act on behalf of human participants, making real-time decisions across the marketplace to optimize outcomes.

Because the protocol is fully on-chain and permissionless, agents can execute **machine-to-machine transactions** autonomously. Providers can earn and clients can consume compute without any direct human intervention. This enables a **fully autonomous compute economy** where agents negotiate, transact, and settle on behalf of their operators.

## For Providers

Providers can run agents that:

* Automatically evaluate incoming tasks
* Calculate profitability and bid strategically
* Optimize resource allocation
* Maximize earnings without manual intervention

Current agent types:

* **Greedy agents**: Bid on all profitable tasks immediately
* **Periodic agents**: Evaluate tasks on a schedule
* **Collaborative agents** (experimental): Providers form groups to coordinate rather than compete with each other

## For Clients

Clients can leverage agents to:

* Optimize task pricing
* Select providers based on performance history
* Monitor and retry failed tasks automatically

## Utility Agents

Beyond bidding and optimization, agents can perform monitoring and maintenance:

* **Security agents**: Scan sources and tasks for malicious patterns
* **Compatibility agents**: Verify hardware and software requirements before registration
* **Health agents**: Monitor provider uptime and performance

## Current Status

Provider agents (greedy and periodic) are operational. Client-side agents are planned.

The vision: a network where agents handle the complexity of pricing, bidding, and optimization while humans simply define their goals. As the ecosystem matures, expect a marketplace of specialized agents that participants can deploy, share, and monetize.

> See [For Providers](/opengpu-network/for-providers/overview.md) and [For Clients](/opengpu-network/for-clients/overview.md) for how agents fit into each workflow.