Demo

This section highlights real-world demonstrations of how Saline SDK enables powerful asset protection through programmable mandates. Each demo includes:

  • A clear use case

  • A YouTube walkthrough

  • GitHub code for replication

These examples go beyond basic scripts. They showcase how Saline protects assets from AI hallucinated or rogue agents, matchers, or bots, ensuring safe automation on and across chains.

Block Assets from Agent

Block Assets from Being Used by Agent

Purpose: Prevent an unauthorized matcher/agent from touching user assets—even if it submits a valid-looking transaction.

💻 Code: GitHub Source

Scenario:

  • User A installs a mandates that restrcit its BTC balance to only stay above 3

  • Matcher A attempts to transfer 1 BTC → ❌ REJECTED

  • Matcher A transfers 2 ETH → ✅ ACCEPTED

Insight: Even if the matcher is allowed to interact generally, fine-grained mandates on assets can override it.

Only Allow Interaction with Whitelisted Agents

Only Allow Interaction with Whitelisted Agents

Purpose: Only trusted agents/LLMs should be allowed to interact with your wallet.

💻 Code: GitHub Source

Scenario:

  • User B creates a mandate that whitelists only Matcher/Agent XYZ.

  • Agent XYZ transfers ETH/BTC from User B address → ✅ ACCEPTED

  • Any other agent attempt to transfer asset from User B → ❌ REJECTED

Insight: Whitelisting agents gives users powerful filters for automation and delegation.

Conditional Swap, Whitelist and Time-frame

Conditional Swap: 1 BTC for >= 20 ETH + Only With ABC Agent + Allow once

Purpose: Showcase conditional, chained logic for automated swaps.

💻 Code: GitHub Source

Scenario:

  • User C creates an intent: “Only swap 1 BTC for ETH >= 20, agent must be ABC and allow once”

  • Matcher ABC proposes 2 BTC → ❌ REJECTED

  • Matcher XYZ proposes 1 BTC for 20 ETH → ❌ REJECTED

  • Matcher ABC proposes 1 BTC for 20 ETH → ✅ ACCEPTED

Insight: Saline lets you specify not just what to trade, but under what logic, with whom and how many time .

LLM Agent With Mandate Awareness (MCP)

Purpose: Demonstrate how a language model connected via MCP can understand and respect user mandates.

💻 Code: GitHub Source

Scenario:

  • User D sets a mandate: “BTC balance must stay >= 2 BTC”

  • LLM Agent attempts 1 BTC transfer → ❌ REJECTED

  • LLM Agent attempts 2 ETH transfer → ✅ ACCEPTED

Insight: Saline makes mandates legible to agents—enabling verifiable AI compliance.

LLM Matcher market-making with MCP

LLM Matcher: Discover + Submit Matching Swaps (MCP)

Purpose: Show how an LLM can act as a decentralized market maker by reading intents, finding matches, and submitting a transaction.

💻 Code: GitHub Source Scenario:

  • LLM queries all swap intents on-chain

  • Finds compatible pairs

  • Submits atomic transaction to execute both sides

Insight: This is decentralized, intelligent market-making. No smart contract needed. Just logic and trustless coordination.

See the Examples folder on GitHub for source code.