Alpaca has unveiled a new open-source resource called the Skills Library, designed to enhance how AI agents interact with its trading infrastructure. Launched on June 17, 2026, this initiative provides a curated set of reusable instructions that enable AI assistants to execute consistent and reliable workflows using Alpaca’s APIs, MCP Server, and command-line tools.
The library addresses a growing need as developers and traders increasingly rely on AI coding assistants for building and testing investment strategies.
By offering standardized guidance and safeguards, it helps ensure that AI-driven processes deliver more predictable and transparent outcomes, moving beyond ad-hoc scripting to structured and repeatable operations.
The Alpaca Skills Library consists of modular workflow packages. Each skill comes as a straightforward Markdown file, known as SKILL.md, containing detailed step-by-step prompts that AI agents can follow.
Users working with compatible tools such as Claude Code, Codex, or Cursor can simply incorporate these skills into their project directories and trigger them via natural language commands.
This approach allows agents to handle complex tasks like strategy development or data analysis while incorporating built-in best practices.
The inaugural offering focuses on backtesting with the Trading API.
This skill guides AI agents through the entire process of turning a vague trading concept into a fully documented research experiment.
It begins by requiring the agent to formalize the strategy, clarifying rules, assumptions, fill models, indicators, and benchmark selections before any simulation runs.
The agent then pulls historical market data via Alpaca’s tools, runs a local backtest, and generates organized outputs including summary reports, trade logs, equity curves, and notes on limitations.
The backtesting skill emphasizes several important elements that strengthen its effectiveness.
It promotes clear strategy formalization to reduce ambiguity in the process. It ensures consistent data retrieval from Alpaca‘s Market Data API, covering bars, quotes, calendars, and corporate actions.
The skill also incorporates standardized performance comparisons against selected benchmarks.
Reproducibility is achieved through saved run directories, raw data, and data fingerprints that allow experiments to be recreated reliably.
In addition, it provides transparent documentation of all assumptions and potential caveats, while supporting optional follow-up validation in paper trading environments.
These features together help transform casual ideas into evidence-based insights that traders can trust and iterate upon. Getting started with the library is straightforward.
Developers clone the official GitHub repository and place the skill folders in the appropriate location for their AI assistant.
For example, Claude Code users add them to the .claude/skills directory, while Cursor supports .cursor/skills or shared .agents/skills directories.
Once set up, the skills integrate easily and can be invoked directly in conversations. Because everything lives in plain Markdown, users can easily review and customize instructions before execution.
Alpaca encourages community involvement, inviting developers to propose new skills, refine existing ones, and share domain-specific expertise.
The company plans to expand the library with additional capabilities around trading execution, market data analysis, research automation, and agent-native features.
This launch builds on recent advancements, including the MCP Server V2 and the Alpaca CLI, which have made it easier to interact with trading and market data endpoints through natural language or terminal commands.
By combining these tools with the Skills Library, Alpaca continues to strengthen its API-first platform for both individual builders and institutional users.
As AI becomes more central to quantitative finance, resources like this Skills Library represent an important step toward more auditable and collaborative automated trading development. The full repository is now accessible on GitHub for immediate exploration and contribution.