3 minutes
The Graph’s AI
Using AI on The Graph
Instead of relying on static datasets or centralized APIs, you can now use our AI-native tooling via the Subgraph MCP, Substreams MCP, and agent skills for both Subgraphs and Substreams.
Why Use Onchain Data with AI?
Using onchain data with AI unlocks powerful new ways to interact with and understand blockchain ecosystems.
AI for Developers
You can use AI to interact directly with The Graph’s data through your agents or build AI-powered applications on top of it. This streamlines development and opens up more intuitive, dynamic use cases.
Enable Natural Language Access to Onchain Data
Model Context Protocol (MCP) servers connect to Claude, Cline, and Cursor. They enable models to understand, query, and interact with structured onchain data using natural language. MCPs remove the need to write low-level queries or interact with APIs directly.
Agent Skills extend AI coding assistants like Claude Code and Cursor with deep, specialized knowledge of The Graph’s protocols. This covers development patterns, best practices, and debugging techniques out of the box.
Subgraph MCP
The Subgraph MCP server connects models to Subgraphs on The Graph Network. It allows language models to explore Subgraph schemas, execute GraphQL queries, find relevant Subgraphs by keyword or contract, and surface usage metrics using natural language.
Benefits of Using Subgraph MCP
- Query Subgraphs and token data using natural language
- Access GraphQL schemas and token metadata without manual queries
- Find Subgraphs by keyword or contract and check 30-day usage
- Retrieve real-time data from The Graph and onchain token sources
Substreams MCP
The Substreams MCP server lets AI agents search, inspect, and analyze Substreams packages — from registry discovery to sink deployment. It supports dual transport for local clients and SSE/HTTP for remote agents.
Tools
| Tool | Description |
|---|---|
search_substreams | Search the substreams.dev package registry by keyword, network, or sort order |
inspect_package | Inspect a .spkg file to see its full module graph, protobuf types, and metadata |
list_package_modules | Lightweight alternative to inspect_package — returns module names, types, and inputs/outputs |
get_sink_config | Analyze a package’s sink configuration and generate ready-to-run CLI commands |
Benefits of Using Substreams MCP
- Discover Substreams packages across any supported chain without leaving your AI assistant
- Inspect module graphs and protobuf schemas using natural language
- Generate deployment CLI commands automatically from sink configurations
Agent Skills for Subgraphs
Subgraph Skills is a collection of AI agent skills that provide expert knowledge for developing, testing, and deploying Subgraphs. Skills are available as Claude Code plugins or OpenClaw skills.
Available Skills
| Skill | What It Covers |
|---|---|
| Subgraph Development | Schema design, manifest configuration, AssemblyScript handlers, contract bindings, Subgraph Composition, and common patterns (ERC20, DEX, NFT, Lending, etc.) |
| Subgraph Optimization | Pruning, @derivedFrom, immutable entities, avoiding eth_calls, timeseries, aggregations, and grafting |
| Subgraph Testing | Matchstick unit testing, the Subgraph Linter, mock events, entity assertions, and CI/CD integration |
Agent Skills for Substreams
Substreams Skills enhance AI coding assistants with specialized Substreams expertise. Install once and your assistant automatically applies Substreams best practices when working on relevant projects.
Available Skills
| Skill | What It Covers |
|---|---|
substreams-dev | Manifest configuration, Rust module types (map, store, index), protobuf schema design, performance optimization, and debugging |
substreams-sql | CDC database sinks, relational mappings, PostgreSQL and ClickHouse patterns, and blockchain data modeling |
substreams-testing | Unit, integration, and performance testing; FireCore tooling; and CI/CD pipelines |