research-llm-apis 2026-04-04

2026-04-05 · Source: Simon Willison's Weblog · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, quick

Summary

A new GitHub repository, "research-llm-apis," was released on April 4, 2026, documenting research into the HTTP APIs of various Large Language Model (LLM) providers. This initiative supports a significant update to the LLM Python library and CLI tool, which currently abstracts over hundreds of LLMs from numerous vendors. The update aims to accommodate new features like server-side tool execution, which the existing abstraction layer cannot handle. The research involved using Claude Code to analyze Python client libraries for Anthropic, OpenAI, Gemini, and Mistral, generating `curl` commands to capture raw JSON outputs for both streaming and non-streaming modes across diverse scenarios. Both the scripts and their captured outputs are now available in the repository.

Key takeaway

For AI Architects designing or updating LLM abstraction layers, you should prioritize direct API research into leading providers like Anthropic, OpenAI, Gemini, and Mistral. Understanding their raw JSON structures and new features, such as server-side tool execution, is critical to building future-proof and extensible frameworks that can adapt to rapid LLM advancements.

Key insights

Analyzing LLM provider APIs directly informs robust abstraction layer design for evolving features.

Principles

• Direct API analysis is crucial.
• Abstraction layers need feature extensibility.

Method

Claude Code analyzed Python client libraries (Anthropic, OpenAI, Gemini, Mistral) to craft `curl` commands, capturing raw JSON for streaming and non-streaming LLM API interactions.

In practice

• Use `curl` for raw API inspection.
• Examine client libraries for API patterns.

Topics

• LLM APIs
• HTTP APIs
• Tool Execution
• Abstraction Layers
• Python Client Libraries

Code references

• simonw/research-llm-apis
• simonw/llm

Best for: AI Architect, AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• LLM Providers & APIs Guide: OpenAI, Claude & Gemini (Models, Endpoints, Usage) — Major LLM providers offer diverse models and APIs for text and image generation.
• Want to Build AI Agents? Start by Calling an LLM API Yourself — Direct LLM API calls are the fundamental building blocks for all AI agents.
• LLM Research Papers: The 2026 List (January to May) — LLM research in early 2026 emphasizes hybrid architectures and long-context efficiency beyond simple scaling.
• Building LLM-Powered Web Apps with Client-Side Technology — Client-side LLM applications offer cost, privacy, and speed benefits by processing data and models locally.
• I think Anthropic and OpenAI have found product-market fit — Frontier AI labs have found product-market fit for coding agents, driving significant enterprise revenue through aggressive pricing.
• Diagnosing Knowledge Gaps in LLM Tool Use: An Agentic Benchmark for Novel API Acquisition — LLM tool use with novel APIs requires specific knowledge components, with usage examples being critical, and retrieval/tuning playing complementary roles.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.