Build a coding agent workflow using raw HTTP calls to LLM APIs.
npx skills add https://github.com/mgratzer/bloomery --skill bloomery
Please help me install this Agent Skill. Skill: bloomery Source: https://github.com/mgratzer/bloomery/tree/main/skills/bloomery Command: npx skills add https://github.com/mgratzer/bloomery --skill bloomery
The Bloomery Skill provides a hands-on guide to building a coding agent from scratch using raw HTTP calls to an LLM API, enabling engineers to learn the mechanics behind agentic loops, memory, and tooling without relying on SDKs.
Run the scaffold script to generate your agent scaffold and begin Step 0.
High-intent search queries and answers about installing and using this skill.
You build a coding agent using raw HTTP calls by implementing a step-by-step workflow with a chat loop, tool definitions, and a dispatcher. This approach scaffolds a working agent in roughly 300 lines of code without relying on SDKs.
Yes, the coding agent tutorial supports OpenAI, Anthropic, and Gemini API endpoints. It guides you through making raw HTTP requests to these compatible LLM providers across TypeScript, Python, Go, and Ruby implementations.
To implement an LLM agent from scratch, you need to build essential tools such as list_files, read_file, and run_bash. These tools are integrated into a multi-step conversational loop managed by a dispatcher with automatic progress detection.
Building an AI agent with raw HTTP requests teaches you the underlying mechanics of agentic loops, memory, and tooling without SDK abstractions. It provides a self-contained workflow that demonstrates end-to-end validation across multiple programming languages.
Yes, you can write a coding agent in Go or Ruby. The tutorial runs across TypeScript, Python, Go, and Ruby, providing step-by-step validation for building a self-contained agent workflow using raw HTTP calls to an LLM API.
The best way to scaffold a multi-step conversational loop is to run the scaffold script to generate your agent structure. From there, you implement the system prompt, tool definitions, and an optional edit tool with step-by-step validation.
binfen1Guide developers to build applications with Claude API and Anthropic SDKs.
copilot-ldOrchestrate LLM completions, memory windows, and tool calls for conversational AI workflows.
i3ringitStandardize agent-native architectures by separating prompts from atomic tools.
bingggCall CloudBase database, functions, and storage over HTTP without SDKs.
