· 4 min read

Introducing AgentKit Build

Introducing AgentKit Build

AgentKit Build makes it possible to create onchain earn products integrated with any leading protocol in an afternoon.

When integrating onchain yields, developers need to choose an integration method, understand transaction handling across chains, confirm endpoint parameters, interpret errors, and keep their implementation aligned with the latest API specification.

AgentKit Build brings that information directly into the developer's AI agent.

It is a standalone skill providing access to integration guides, API references, chain-specific transaction patterns, documentation search, and the live OpenAPI specification. When live API data is needed, the skill connects to the Yield.xyz MCP server automatically. Developers can get implementation guidance without leaving Claude Code or another AI development environment.

One infrastructure layer for onchain yield

Neobanks, wallets, and custodians are adding onchain earn to savings products, and AI agents are beginning to manage DeFi positions autonomously.

Yield.xyz supports these products through a single API for discovering opportunities, constructing transactions, and tracking positions across more than 3,000 onchain opportunities and 80+ networks.

The infrastructure is used by teams including Ledger, Trust Wallet, Anchorage Digital, Turnkey, Crossmint, Utila, and Deblock, with products reaching more than 250 million users. AgentKit made Yield.xyz's infrastructure callable by AI agents. And now, AgentKit Build gives developers the information required to build on top of it.

What AgentKit Build unlocks

AgentKit Build is a standalone skill available through Claude and other AI models.

There is no separate authentication flow or additional server to configure. When live API access is needed, the skill connects to the Yield.xyz MCP server automatically.

Bundled integration references

Core implementation guidance is compiled directly into the skill, including:

AgentKit Build indexes the complete Yield.xyz documentation set.

Developers can search API references, product guides, endpoint behavior, and implementation patterns from inside their agent. Results are ranked by relevance, removing the need to switch between an editor, browser, and documentation search.

Live OpenAPI introspection

AgentKit Build can query the deployed Yield.xyz OpenAPI specification on demand. This provides current field definitions, required flags, request formats, response schemas, and endpoint-specific error structures. The agent can verify implementation details against the production API rather than relying only on static documentation.

Choose the right integration path

AgentKit Build searches the live documentation index and returns ranked links to the most relevant integration guides — SDK, REST API, or Widget. Reading those pages surfaces the technical and operational reasoning behind each approach.

Handle transactions across 12 chain families

Transaction construction and signing vary across Ethereum, Solana, Cosmos, Tron, and other supported networks.

AgentKit Build pairs two tools to surface chain-specific guidance. The skill searches the live documentation index for the chain you're integrating and returns ranked page links. It then returns the full markdown for the page you pick, covering unsigned transaction formats, required arguments, signing SDKs, broadcast requirements, and implementation patterns.

The skill itself bundles dedicated references for all 12 supported chain families (Aptos, Cardano, Cosmos, EVM, NEAR, Solana, Stellar, Substrate, Sui, Tezos, TON, and Tron), so developers can get answers regarding chain-specific questions.

Implement yield discovery

The skill returns the endpoint's complete integration surface, including filters, enum values, sort options, pagination behavior, and response fields. Developers can use it to implement queries by network, asset, yield type, or other supported parameters.

Diagnose API errors

AgentKit Build retrieves the authoritative error shape and reasons about it against developer-supplied context (endpoint, arguments). This helps identify missing fields, invalid parameters, unsupported values, and incorrect transaction handling from the same environment where the integration is being written.

AgentKit and AgentKit Build

AgentKit and AgentKit Build support different parts of the product lifecycle.

AgentKit provides runtime actions. Agents can discover opportunities, evaluate metadata, construct transactions, and track positions across supported protocols and networks.

AgentKit Build provides development intelligence. Developers can inspect the API, search documentation, select an integration method, understand chain-specific requirements, and diagnose implementation issues.

A developer can use AgentKit Build to understand the required architecture, then use AgentKit's action tools while implementing and testing the product. The same Yield.xyz infrastructure supports the integration process and serves the product after launch.

Build without leaving your agent

AgentKit Build reduces the time between starting an integration and constructing the first working transaction. Developers can ask questions such as:

The agent can search the documentation, retrieve the relevant guide, and verify details against the live API specification.

Get started

AgentKit Build is available as a Claude Code plugin or as a standalone skill in any AI agent.

Claude Code (recommended) — installs the skill and auto-connects the live doc tools in one step:

/plugin marketplace add stakekit/agentkit
/plugin install yield-xyz-agentkit-builder@agentkit

Any other agent (Codex, Gemini CLI, …) — install the standalone skill:

npx skills add <https://github.com/stakekit/agentkit>

No API key is required to use the skill's documentation tools — you'll need your own Yield.xyz API key once you start building an integration.

Documentation: https://docs.yield.xyz/docs/yieldxyz-agentkit-builder-skill

GitHub: github.com/stakekit/agentkit