Skip to main content
Exponent’s documentation is structured for accurate consumption by both AI agents and humans. Without current, structured documentation, AI coding assistants produce incorrect API signatures — particularly problematic for Solana transactions where parameter types must be exact. Exponent provides four machine-readable files that give agents exact method signatures, parameter types, and usage patterns — verified against the SDK source.

​
Agent Files

llms.txt

Structured index following the llms.txt standard — package info, program IDs, and links to every documentation section

llms-full.txt

Complete API reference with every method signature, parameter type, and return type across all SDK classes

skill.md

Operational knowledge file — setup pattern, decision tree, constraints, and common mistakes

AGENTS.md

Documentation editing instructions — terminology glossary, page templates, import rules, and style guide

​
What Each File Does

​
llms.txt

A structured index file following the llms.txt standard. This is the first file an agent should fetch when working with Exponent. Contains: package name, install command, all Solana program IDs, links to every documentation section, key classes and their load patterns, supported yield source flavors. How agents use it: When an AI tool sees @exponent-labs/exponent-sdk in your codebase, it fetches llms.txt to understand what the SDK does and where to find details.

​
llms-full.txt

The complete API reference — every public method on every class, with exact parameter names, types, and return types. Contains: core concepts and formulas, every method on Vault, MarketThree, Orderbook, Router, YtPosition, LpPosition, full parameter signatures, return types, environment configuration, and utility exports. When to use: When you need an agent to write a specific SDK call and want it to get the types right.

​
skill.md

An operational knowledge file that turns an AI agent into a competent Exponent SDK user. Unlike llms.txt (an index) and llms-full.txt (a reference), skill.md teaches the agent how to think about Exponent. Contains: setup pattern (exact imports and initialization code), decision tree mapping goals to classes and methods, constraints (CU limits, ALT requirements, YieldPosition initialization), common mistakes and how to avoid them.

​
AGENTS.md

Editorial instructions for AI agents editing the documentation itself. If you’re using an AI tool to contribute to or maintain the Exponent docs, this file ensures consistency. Contains: source of truth paths, exact terminology glossary, page templates for instructions/read functions/account references, import rules, style rules, and Mintlify component usage.

​
Using with AI Tools

​
Cursor

Add skill.md or llms-full.txt as a project-level doc in Cursor settings. Cursor will use it to generate correct SDK calls when you write Exponent code.

​
VS Code Copilot

Download llms-full.txt to your project root. Copilot automatically includes workspace files in its context, so SDK method signatures will be accurate.

​
Claude / ChatGPT

Paste the contents of skill.md into your conversation, or point the agent at the URL. The decision tree and constraints give the model everything it needs to write correct Exponent transactions.

​
Best Practices

  1. Point your AI tool at skill.md before starting an integration. This gives it the decision tree and constraints.
  2. Use llms-full.txt for reference calls. When the agent needs exact parameter types, this file has them.
  3. Verify against the docs. Agent output should match what’s documented on the instruction pages.
  4. Add skill.md to your repo as a reference file. All team members’ AI tools will pick it up.
  5. Pin the SDK version. llms-full.txt documents the current @exponent-labs/exponent-sdk API. If you use a different version, parameter types may differ.