Mimir

Mimir is an advanced autonomous blockchain analysis agent built for the ASI Agentverse platform. You can tag @mimir in ASI:ONE chat to ask questions about blockchain activity, and it will investigate and explain what’s happening on-chain in real time.

Mimir integrates the Blockscout MCP (Model Context Protocol) to access verified blockchain explorer data, allowing it to analyze transactions, token transfers, contracts, and smart token behaviors. It automatically infers the correct blockchain (e.g., Ethereum, Polygon, Arbitrum, Base, BSC) and fills in required parameters like chain_id.

Powered by ASI’s asi1-extended LLM model, Mimir reasons through each request step by step — planning, retrieving data, and producing human-readable summaries. It starts every session with an unlock call (unlock_blockchain_analysis) that provides domain-specific instructions to guide its analysis.

Key Features:

LLM + MCP Integration: Combines AI reasoning with real blockchain data for precise insights.

Auto Parameter Handling: Infers and fills missing values like chain_id and time ranges.

Pagination & Historical Search: Follows pagination automatically and can narrow time windows to find past events.

Resilient & Secure: Async-safe design with retries, backoff, and masked logs.

Self-Updating: Refreshes tool lists and API keys every 12 hours.

Smart Context Management: Keeps essential reasoning while trimming conversation history.

Use Cases:

Analyze transactions or contracts to explain on-chain activity.

Identify token standards (ERC20, ERC721, etc.) and contract verification.

Detect unusual transfers, ownership changes, or suspicious patterns.

In short, Mimir is an autonomous blockchain analyst that connects powerful AI reasoning with real-time blockchain data, turning complex on-chain information into clear, actionable insights.

We used the ASI's Agentverse for hosting of Mimir as it makes him available across the Agentverse, and other users can simply tag him @mimir in their chats. To access on-chain data, we leveraged Blockscout's MCP. The user message that Mimir receives is sent to an ASI LLM to infer which MCP tools will be required and what the user's intent is. Afterwards, we run the required tools sequentially, and as the last step, we send the tool responses and the intent-based prompt to an ASI LLM to generate the final response for the user.