Floster

Floster is a visual, AI assisted workflow automation builder for Chainlink CRE. It solves the problem that building CRE workflows today requires you to manually write TypeScript, learn the @chainlink/cre-sdk API, author YAML config files, and understand the full CRE execution model 

1. CRE workflow authoring is difficult for many : Building a Chainlink CRE workflow requires writing @chainlink/cre-sdk TypeScript by hand, knowing which capabilities exist, and wiring triggers/handlers correctly. Floster generates all of that code from a visual canvas by utilising the CRE as core engine.
2. No feedback loop during development : Without Floster, you write code, run CRE commands and debug raw CLI output in a separate terminal. Floster wraps the entire compile > simulate > output pipeline inside the UI so you see results instantly.
3. High barrier for non experts : Describing a workflow in plain English shouldn't require knowing the CRE workflow schema. Floster's AI layer turns a natural language prompt directly into a valid, runnable CRE workflow.
4. Broken workflows are hard to fix : The AI layer doesn't just generate workflows, it also repairs them, so if a workflow fails validation, you can ask the AI to fix it rather than debugging the JSON manually.

• Frontend : React 19 + Tailwind CSS, drag-and-drop canvas where nodes and edges map directly to CRE triggers and actions
• Backend : Node.js + Express server that exposes /validate, /compile, /simulate, and /ai REST endpoints
• Workflow compiler : A pure TypeScript code generator (templates.ts) that takes the workflow JSON and emits ready-to-run @chainlink/cre-sdk TypeScript + all required config files
• Simulation engine : Shells out to the cre CLI binary, bootstraps the Javy/Wasm plugin, and streams cre workflow simulate output back to the UI in real time
• Validation & lint : JSON Schema validation + a custom determinism linter that catches bad configs before compilation
• AI layer : Calls LLM with a detailed system prompt that knows the full CRE workflow schema, generates or repairs workflows from plain English
• Smart contracts : Solidity on Ethereum Sepolia, built and tested with Foundry, implements the CRE IReceiver interface to execute on chain ERC20 transfers triggered by the Chainlink Forwarder
• Chainlink CRE as the execution backbone : CRE is not just a target, it's what makes the whole thing work.

1. The compiler generates TypeScript that imports @chainlink/cre-sdk and uses cre.capabilities.HTTPClient for HTTP actions, 
2. cre.capabilities.EVMClient for on-chain reads/writes and ERC20 transfers, 
3. ConfidentialHTTPClient for TEE-secured API calls, and 
4. CronCapability / HTTPCapability / logTrigger for all three trigger types.
5. Every workflow is wired together using cre.handler.