Watch on-chain events with SQL, deliver alerts to Webhooks, Redis, or Kafka with sub-second latency. Reorg-safe, idempotent, and built on Arrow + DataFusion.
SELECT t.hash AS tx_hash, l.address AS contract, l.topic0 AS event_sig, l.data AS raw_data FROM main.logs l JOIN main.transactions t ON t.hash = l.transaction_hash WHERE l.topic0 = '0xddf252ad...' -- ERC-20 Transfer AND CAST(l.data AS BIGINT) > 1000000 -- > 1M tokens ORDER BY t.block_number DESC LIMIT 10
Query blocks, transactions, and decoded logs in real-time using standard SQL. Powered by Apache Arrow and DataFusion for columnar speed.
Route matched events to Webhooks, Redis Streams, or Kafka topics. Built-in retry logic with configurable dead-letter queues.
Idempotent delivery keys and automatic retraction on chain reorganizations. Never process a reverted event twice.
Account-based workspaces with OAuth login, project-scoped API keys, and role-based access for teams and MCP agents.
Upload ABI JSON files and query decoded event parameters directly. Supports ERC-20, ERC-721, Aave, Uniswap, and custom ABIs.
AI agents and copilots can query live blockchain data, create watches, and inspect delivery pipelines through standard MCP tool calls.
Ever Green ships a native MCP client that works with Claude Desktop, Cursor, and any MCP-compatible host. Install the evergreen-mcp binary or use the Docker image — no repo access required.
Sign in at onchainsql.xyz, open API Keys, and create a project key with viewer or editor role.
Download the evergreen-mcp binary for macOS or Linux, or pull the Docker image from ghcr.io/msbeni/evergreen-mcp.
Add evergreen-mcp to your Claude Desktop or Cursor MCP config with your API key and the base URL https://api.onchainsql.xyz.
Your AI agent can now run SQL queries, create watch operators, inspect delivery records, and manage your blockchain monitoring pipeline.
Connect to any EVM-compatible RPC endpoint. Blocks, transactions, and logs are streamed into an in-memory Arrow columnar store with configurable retention.
SQL watch operators run against every new block. DataFusion executes queries over Parquet-backed data with sub-millisecond latency on local hardware.
Matched events are queued in an outbox and delivered via your configured target with idempotency keys, retry policies, and dead-letter handling.
When the chain reorganizes, delivered events from orphaned blocks are automatically retracted. Pending outbox entries for reverted blocks are cancelled.
Execute SQL against live blocks, transactions, and decoded logs. Returns JSON with columns, rows, and timing metadata.
List, create, enable, disable, and delete SQL watch operators with webhook, Redis, or Kafka delivery targets.
Discover available tables (main.blocks, main.transactions, main.logs) and describe their columns and types.
Inspect recent delivery events with matched row samples, dispositions, retry counts, and dead-letter status.
Project-scoped API keys with admin, editor, and viewer roles. OAuth 2.0 via GitHub and Google. Magic-link sign-in.
Production-grade observability with delivery telemetry, query stats, auth counters, and block markers at /metrics.
Deploy Ever Green alongside your node, write SQL watch queries, and let the watchtower handle delivery, retries, and reorg safety.
Get Started Free