Skills vs MCP
LaserSell offers two AI integration paths. They serve different purposes and work best together.| MCP Server | Coding Skills | |
|---|---|---|
| How it works | Searches LaserSell documentation on demand when the assistant needs to look something up | Loads deep SDK knowledge directly into the assistant’s context before it starts writing code |
| Best for | Looking up API schemas, finding specific docs pages, answering questions about LaserSell | Writing SDK integrations, building transactions, connecting to the stream, handling errors |
| When knowledge is available | On demand (the assistant must ask for it) | Immediately (the assistant already has it in context) |
| Setup | Add MCP server config to your editor | Install the skill package |
What’s Included
The LaserSell coding skill ships a structured knowledge tree that the assistant loads when working on LaserSell code:- SKILL.md — root skill file with an overview of the SDK, key concepts, and pointers to detailed references
- Working examples in 4 languages — TypeScript, Python, Rust, and Go examples covering buy/sell transactions, stream connections, strategy configuration, and transaction submission
- Strategy guide — detailed reference for target profit, stop loss, trailing stop, sell on graduation, and deadline timeout configuration
- Troubleshooting guide — common errors, their causes, and fixes
- API and type references — request/response schemas, enum values, and field descriptions for every endpoint
- Stream protocol reference — the full Exit Intelligence Stream message format, event types, and connection lifecycle
- Templates — starter code templates for common integration patterns
Install
Option 1: npx (Recommended)
Option 2: Claude Code CLI
Option 3: Manual
Copy theskills/lasersell/ directory from the LaserSell skills repository into your project’s .claude/skills/ folder:
What the Skill Covers
Once installed, your AI assistant has full context on these topics: Building buy and sell transactions. The assistant knows how to constructBuildBuyTxRequest and BuildSellTxRequest objects, call the LaserSell API, and handle the response across all four SDK languages. It understands output types (SOL vs USD1), slippage configuration, and partial sells.
Exit Intelligence Stream. The assistant can write stream connection code, configure strategy parameters, handle all server event types (pnl_update, exit_signal_with_tx, position_opened, position_closed, liquidity_snapshot), and process exit signals end to end.
Partial and multi-leg sells. The assistant understands how to query slippage bands, determine the maximum sellable amount at a given slippage threshold, and build partial sell transactions. It can also construct multi-leg sell sequences for large positions.
Liquidity snapshots. The assistant knows how to read slippage bands, interpret liquidity trends (growing, stable, draining), and use snapshot data to size sells optimally.
Transaction signing and submission. The assistant can write code for local transaction signing, and understands all send targets: default RPC, custom RPC, and Helius Sender. It knows when to use each and how to configure them.
Error handling. The assistant has context on common error codes, retry patterns, slippage failures, and connection recovery for both the REST API and the stream.
Combining Skills and MCP
For the best AI coding experience, use both the coding skill and the MCP server together:- Install the coding skill so your assistant has deep SDK knowledge loaded into context whenever you write LaserSell code.
- Connect the MCP server so your assistant can search the full documentation when it needs to look up something outside the skill’s scope, like partner fee details, rate limit tiers, or changelog entries.

