Axiory has launched a platform designed from the ground up for AI agents to trade global financial markets — not humans using dashboards, but agents accessing markets directly through a standardized interface.
The platform provides access to spot CFDs including FX, stocks, and ETFs through a single infrastructure layer. Agents can retrieve market data and execute structured trading instructions autonomously.
Four-Layer Architecture
Axiory’s system is built around four layers that accommodate both human developers and autonomous agents:
- Traditional Trading API — Standard programmatic market access
- Cloud Code Environment — AI workflow execution in the cloud
- MCP Server — Direct AI agent connectivity via Anthropic’s Model Context Protocol
- Strategy Builder (forthcoming) — Converts natural language trading ideas into fully autonomous strategies running in the cloud
The MCP integration is notable. While MCP has seen rapid adoption across Google, OpenAI, and Microsoft for connecting AI to external tools, its application to financial markets has been limited — typically focused on single asset classes like US equities or crypto derivatives. Axiory aims to be the first multi-asset MCP trading layer.
The Agent-Native Thesis
“We expect a shift away from traditional trading interfaces toward personal AI assistants executing trades on behalf of users,” said David Kašper, director of Purple Technology and co-founder of Purple Group, which developed the platform’s technology. “This is a direction the financial technology sector is already moving toward.”
CEO Roberto d’Ambrosio framed the goal as infrastructure consolidation: allowing agents and traders to access multiple markets from one environment rather than through fragmented systems.
The platform launches with a demo environment for testing AI agents and strategies using virtual funds — a sandbox approach that mirrors how enterprises are deploying agents in other domains.
Context: Agents Enter Finance
Axiory joins a growing wave of agent-native financial infrastructure:
- KX launched agentic AI blueprints for capital markets at Nvidia GTC, with RBC proof-of-concept
- Agentic commerce is enabling agents to make real payments
- OpenAI’s Frontier platform partnered with McKinsey and BCG for enterprise agent workflows including financial analysis
The pattern: every industry vertical that involves data retrieval, decision-making, and execution is getting an agent-native layer. Finance is a natural fit — high-frequency decisions, structured data, clear execution targets.
The Risk Question
Agent-driven trading raises obvious governance questions. When an autonomous agent executes trades based on natural language strategies, who is liable for losses? How do existing regulatory frameworks (MiFID II, SEC rules) apply to non-human traders?
These are the same questions Google’s cross-enterprise agent framework raised yesterday: trust, permissions, and governance for agents that operate across organizational boundaries with real-world financial consequences.
The demo environment with virtual funds is a pragmatic first step. But the path from sandbox to live markets will require regulatory clarity that doesn’t yet exist.
From healthcare agents to recruiting agents to trading agents: the vertical expansion of agentic AI continues. Each new domain raises the same question — who controls the agent when the stakes are real?