TL;DR
- Agent Infrastructure: GMX rebuilt its documentation, SDK, and API to support autonomous agents with consistent schemas, typed functions, and structured data.
- Composable Skills: Pre-built skills let agents perform complex GMX operations such as opening positions or adjusting risk with minimal configuration.
- Real Use Case: Elfa AI uses the upgraded stack to move from signal generation to automated execution on GMX, creating a continuous research-to-trade workflow.
Autonomous AI agents are quickly becoming active participants in on-chain markets, moving far beyond simple bots and into systems capable of multi-step reasoning, cross-protocol execution, and continuous risk assessment. As these agents take on more responsibility, they require infrastructure built specifically for programmatic consumption. GMX is positioning itself at the center of this shift with a fully reworked integration layer designed to support agent-driven activity at scale.
A Documentation Overhaul Built for Programmatic Consumers
Most DeFi protocols were never designed with autonomous agents in mind, and inconsistent documentation often becomes a hidden bottleneck. GMX rebuilt its documentation from the ground up to eliminate ambiguity and provide the structured clarity agents need. Function references now follow consistent naming conventions, parameters include explicit types and valid ranges, and return schemas are fully detailed. Code examples are crafted to serve both human developers and language models, reducing the risk of errors or hallucinations during execution. The goal is simple: ensure both a developer and their agent can rely on the same source of truth.
Typed SDK, Structured API, and Composable Skills
The GMX SDK introduces a typed, structured interface that removes the need for raw contract calls or manual ABI management. Because the SDK exposes predictable return shapes, developers can wrap it directly into agent tools without heavy engineering. The GMX API follows the same philosophy, offering schema-stable endpoints for real-time market data, position reads and writes, fee estimation, and historical datasets for backtesting.
Beyond the API, GMX has invested in a growing library of agent skills: pre-built modules that handle common operations such as opening leveraged positions, setting stop-losses, or reading funding rates. These skills are composable, allowing agents to chain actions into multi-step strategies that would otherwise require significant custom development.
Early Production Use: Elfa AI’s Autonomous Trading Loop
Elfa AI is among the first teams to deploy the upgraded stack in production. Its system continuously indexes price movements, social chatter, news, and prediction markets to surface actionable insights. Elfa Auto, the team’s autonomous trading agent, completes the loop by executing trades directly on GMX when user-defined conditions are met. Research flows into execution without interruption, enabled by the SDK, API, and skills working together. It is a capability neither team could deliver alone, and it highlights how GMX’s integration layer unlocks new agent-driven workflows.
