Caroline Bishop
Mar 13, 2026 16:40
Dune ships CLI and Skills framework giving AI agents direct terminal access to blockchain data across 130+ chains without human intervention.
Dune Analytics has released a command-line interface and Skills framework that allows AI agents to autonomously query onchain data across more than 130 blockchains. The March 13 launch follows last week’s Model Context Protocol (MCP) release, completing what the $1 billion-valued analytics platform calls its “agent-native” infrastructure.
The move positions Dune for a shift in how blockchain data gets consumed. Forget dashboards—autonomous agents running in terminals now need direct programmatic access to decoded smart contract data, DEX trades, and protocol metrics.
What Actually Shipped
The CLI, built on a Go SDK, handles the full analyst workflow: dataset discovery, DuneSQL execution, query management, and usage tracking. Every output comes as JSON, designed for machine parsing rather than human reading.
More interesting is the Skills.md file bundled with the CLI. This follows an emerging standard adopted by Anthropic, Microsoft, and others—essentially a capability manifest that tells AI agents what Dune can do, which commands to run, and how to handle errors. An agent reads the file and immediately understands the toolset without custom integration work.
Two new API endpoints power both the CLI and MCP under the hood. The /v1/datasets/search endpoint lets agents search Dune’s entire catalog using natural language with filters for chain, category, and schema. The /v1/datasets/search-by-contract endpoint takes any contract address and returns every decoded event and function table associated with it.
Why Terminal Access Matters
Terminal-based AI tools like Claude Code, Cursor, and various autonomous agent frameworks operate in environments where every token counts. CLI plus Skills offers lower overhead than persistent MCP connections for these workflows.
The practical implication: an AI agent can discover relevant tables, write a query, execute it, and return results without human involvement. For trading bots, research automation, and monitoring systems, that’s a meaningful capability upgrade.
Context and Competition
Dune has expanded aggressively since hitting unicorn status in February 2022 after a $69.42 million Series B led by Coatue. The platform now supports data from over 130 chains, up from its Ethereum-only origins. In February 2025, the company announced reaching 100 supported blockchains as multichain usage accelerated.
The agent-native push addresses a real infrastructure gap. As AI agents proliferate in crypto—handling everything from portfolio rebalancing to governance participation—they need reliable data pipelines that don’t require human intermediaries clicking through web interfaces.
Getting Started
Developers can install the CLI and authenticate with an API key to start querying immediately. The choice between MCP and CLI depends on the use case: MCP suits GUI-based AI applications like Claude.ai or ChatGPT with persistent connections, while CLI plus Skills fits terminal agents and autonomous workflows better due to stateless operation and lower token overhead.
Whether this becomes standard infrastructure for crypto AI agents depends largely on adoption by the autonomous agent frameworks now proliferating across the ecosystem. The tooling exists—now it needs builders to actually use it.
Image source: Shutterstock
Credit: Source link
