Skip to main content

For AI Agents & LLMs

Push Chain provides structured, machine-readable resources so AI coding assistants, agents, and automation pipelines can understand and execute cross-chain actions with maximum reliability.

Context Files​

AI models have a context window — the amount of text they can process at once. Providing structured documentation upfront helps models give precise answers without hallucinating.

Push Chain offers two context files:

FileBest for
/llms.txtCompact summary with links to every resource. Works with most models (100K+ token context).
/llms-full.txtFull documentation corpus inline. Use when your model has a large context window or you want deep reference without following links.

Add to Your AI Code Editor​

Cursor​

  1. Open Cursor Settings → Features → Docs
  2. Click Add new doc and paste one of the following:
https://push.org/llms.txt
https://push.org/llms-full.txt
  1. Use @Docs → Push Chain in the chat to reference Push Chain documentation.

Windsurf​

Add to the Cascade window (CMD+L) at the start of your conversation:

@docs:https://push.org/llms.txt
@docs:https://push.org/llms-full.txt

Claude Code​

Reference directly in your prompt or CLAUDE.md system context:

https://push.org/llms.txt

For richer integration, use the structured /agents/ layer below — it provides typed capabilities, execution workflows, and decision trees that go beyond static documentation.

Zed / Other Editors​

Paste the URL into your AI assistant's context or system prompt. The /llms.txt format is understood by any LLM.

Agent Layer (/agents/)​

Push Chain goes beyond static documentation with a full machine-readable execution layer at /agents/. This is organized as a layered stack: discovery → capabilities → execution → validation.

FileWhat it contains
/agents/index.jsonDiscovery map — every file, its purpose, and the recommended traversal order
/agents/capabilities.jsonEvery SDK capability with inputs, outputs, and method signatures
/agents/sdk-capabilities.jsonFull SDK namespace map including all methods and advanced arguments
/agents/supported-chains.jsonVerified chain list with CAIP-2 IDs, RPC URLs, explorers, and contract addresses
/agents/contract-addresses.jsonVerified smart contract addresses for Push Chain core contracts, PRC-20 tokens, AMM pools, and all external chain gateways
/agents/workflows/index.jsonStep-by-step execution guides for all common tasks
/agents/schemas/index.jsonJSON schemas for all SDK request/response types
/agents/decision-tree.jsonBranching logic to select the right capability from user intent
/agents/task-router.mdPlain-language routing guide mapping goals to capabilities
/agents/errors.jsonError catalog with recovery actions for every known failure mode
/agents/retrieval-map.jsonMaps every capability to its authoritative documentation source (for RAG)

Canonical Workflows​

Ready-to-execute step-by-step guides for the most common Push Chain tasks:

WorkflowWhat it does
/agents/workflows/initialize-client.mdCreate a PushChainClient from any signer
/agents/workflows/create-universal-signer.mdWrap an EVM or Solana signer into a UniversalSigner
/agents/workflows/send-universal-transaction.mdExecute a transaction on Push Chain from any origin chain
/agents/workflows/send-multichain-transaction.mdSend to an external chain via CEA or cascade pattern
/agents/workflows/track-transaction.mdMonitor universal transaction lifecycle
/agents/workflows/sign-universal-message.mdSign a message for off-chain verification
/agents/workflows/read-blockchain-state.mdQuery on-chain state via EVM clients
/agents/workflows/use-contract-helpers.mdInteract with UEA Factory and other native contracts
/agents/workflows/constants-reference.mdAll chain IDs, token constants, and SDK enums
/agents/workflows/configure-dev-environment.mdInstall SDK and configure Hardhat / Foundry / Remix

Integration Paths​

Human developer​

  1. Follow the Quickstart to run your first transaction
  2. Work through Tutorials for end-to-end flows
  3. Use the SDK and UI Kit for production integration

AI agent / copilot​

  1. Fetch /llms.txt as the entry layer
  2. Load /agents/index.json to discover all capabilities and workflows
  3. Use /agents/decision-tree.json to map user intent to the right capability
  4. Execute using the matching workflow from /agents/workflows/
  5. Validate with schemas from /agents/schemas/
  6. Fetch /llms-full.txt when full inline context is needed

RAG / retrieval pipeline​

  1. Index /llms-full.txt as your document corpus
  2. Use /agents/retrieval-map.json to map queries to authoritative sources
  3. Ground responses with canonical workflows from /agents/workflows/

Notes​