Modulus launches desktop app to share memory across coding agents

Modulus released a desktop application that lets developers run several AI coding agents while keeping a shared memory of the entire codebase, eliminating the need to re‑explain backend APIs when switching between repos [1]. The founders, Jeet and Husain, built the tool after repeatedly copying API schemas between Cursor windows while working on a startup that spanned a backend and a frontend repository. Their custom Memory and Context Engine now synchronises knowledge across any number of repositories, from backend services to shared libraries and AI‑agent code.

By centralising context, Modulus aims to replace ad‑hoc solutions like Conductor and address two pain points: agents losing context when switched and the inability of agents to understand dependencies across multiple repositories. The app integrates with Cursor via the MCP protocol, allowing seamless hand‑offs between agents without manual prompts. The team plans to open an API so teams can swap coding agents or IDEs without losing state, eventually supporting collaborative AI‑driven workflows.

A short demo is available via a launch post on X, and the founders are seeking feedback from the HN community [1]. If adopted, Modulus could streamline multi‑repo development cycles, reducing the overhead of context duplication and potentially accelerating AI‑assisted coding across larger codebases. The open question remains whether developers will prefer a dedicated desktop layer over existing IDE plugins or whether the added complexity of a separate memory engine will be justified in practice.