Modulus launched this week as a macOS desktop app aimed at developers running multiple AI coding agents at once. The friction it targets is straightforward: when working with tools like <a href="/news/2026-03-14-anthropic-silent-ab-test-claude-code">Claude Code</a> across several repositories, developers typically hand-feed each new agent session the relevant context — API schemas, dependency graphs, recent code changes — before any work can begin. Modulus automates that through a shared "Project Memory" layer that pushes cross-repository knowledge to every active agent session automatically, replacing what the company describes as "copy-pasting READMEs into chat windows."
Workspace isolation runs on git worktrees, a native Git feature that lets multiple working trees be checked out from the same repository simultaneously. Each agent gets its own worktree, so a developer can run one agent fixing a bug and another building a new feature in the same codebase without the two stepping on each other. A unified review interface consolidates all agent-generated diffs in one place, letting developers inspect changes and <a href="/news/2026-03-14-metr-research-half-of-swe-bench-passing-ai-prs-rejected-by-maintainers">open pull requests</a> without jumping between terminal windows or IDE instances.
Claude Code is listed as a supported runtime, though Modulus does not claim an official integration with Anthropic — it's a compatibility claim, not a partnership. The app is currently free. A Discord community and a macOS-first rollout with Windows and Linux on a waitlist are a scrappy early-stage launch by any measure. No founding team or organizational background appears on the company's website, and the Hacker News "Show HN" thread offered limited additional technical detail. That leaves the precise mechanics of the Project Memory engine — how it identifies what context is relevant, how it keeps that context current — as the product's most substantive open question.