DeepSeek-Coder open foundation models for code
Open-source code foundation models trained on project-level code with 16k context windows and specialized fill-in-the-middle code synthesis.
Image: GitHub
The weekly · Week 13 · 23 Mar – 29 Mar 2026
AI coding agents: what changed
Developer tooling takes center stage this week with the open-weights DeepSeek-Coder models, advanced caching and tooling in the Anthropic Python SDK, the mcp-cli inspection utility, and Phoenix observability for AI agents.
4 picks
Anthropic Python SDK with prompt caching and tool calling
Official Python SDK for Anthropic Claude models supporting structured output schemas, streaming agent completions, and prompt caching.
Image: GitHub
mcp-cli: Command-Line Inspector and Tester for MCP Servers
An interactive CLI utility to test, inspect schemas, and invoke tools across local and remote Model Context Protocol servers.
Image: GitHub
Phoenix: AI Observability & Agent Tracing Platform
Open-source tracing and evaluation platform that visualizes multi-step agent trajectories, tool calls, and LLM latency.
Image: GitHub
Picked from public sources (GitHub, Hacker News, DEV, Lobsters, Reddit, Product Hunt) by our daily scout. Each summary uses only what the source itself says; GitHub stars are as checked on the date shown.