Univer: The Office Harness for AI Coding Agents
An open-source office engine uniting spreadsheets, documents, slides, and relational tables into a single runtime built for programmatic agent manipulation.
Image: GitHub
The weekly · Week 38 · 14 Sept – 20 Sept 2026
Integration and automation take center stage with Univer's agent harnesses, Hindsight's self-improving memory, and Claude Code Action bringing agents directly into CI/CD pipelines.
4 picks
An open-source office engine uniting spreadsheets, documents, slides, and relational tables into a single runtime built for programmatic agent manipulation.
Image: GitHub
Persistent memory framework that learns from conversational trajectories, ranks historical context, and updates long-term procedural memory between sessions.
Image: GitHub
A self-hosted control plane to orchestrate, audit, and dispatch AI coding agents across engineering teams. It enforces fine-grained access policies and workflow visibility.
Image: GitHub
Official GitHub Action that executes Claude Code on pull requests to reproduce test failures and draft fixes. It also performs continuous security verifications.
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.