Skip to content

AI Assistant

Ask questions in natural language and let AI generate graph queries, analyze patterns, and surface insights.

Overview

The AI Assistant runs inside every investigation. Click AI in the toolbar to open the chat panel. Ask questions about the addresses in your investigation, and the AI will generate Cypher or SQL queries, execute them, and present the results.

The AI is powered by a self-hosted LLM via Ollama, ensuring your investigation data never leaves the ChainGraph infrastructure.

What you can ask

  • “Show all addresses connected to this wallet”
  • “Find the largest transfers in the last 1000 blocks”
  • “Is this address connected to any known exchanges?”
  • “What tokens has this address traded?”
  • “Find circular flows involving this address”
  • “Summarize the activity of this address over the past month”
  • “Which DEX protocols has this address used?”

How it works

  1. You type a natural language question
  2. The AI analyzes your question and the current investigation context
  3. It generates a Cypher or SQL query (shown to you for review)
  4. You can edit the query or execute it directly
  5. Results are displayed as a table and optionally added to the graph canvas
  6. The AI provides a natural language summary of the results

Context awareness

The AI has access to your investigation context — the addresses, nodes, edges, and annotations currently on the canvas. This means it can answer questions relative to your current investigation state.

Query review

Every AI-generated query is shown to you before execution. You can:

  • Execute — run the query as-is
  • Edit — modify the query in the editor before running
  • Copy — copy the query to use in Power Mode
  • Reject — discard the query and ask a different question

Rate limits

TierAI messages/day
Free5
Pro200
EnterpriseUnlimited

Tips

  • Be specific — “Find transfers over 10 ETH from 0xabc” works better than “Show me big transfers”
  • The AI understands blockchain terminology — use terms like “DEX swap”, “ERC-20”, “internal call”
  • For complex analysis, break it into multiple questions
  • Review generated queries to learn Cypher and SQL patterns
AI Assistant — ChainGraph Docs