The Spotify Engineering team unveiled a multi-platform environment called Xirp, built specifically for managing AI coding agents. Inside Spotify, more than 1,300 engineers already use it, and now it's available in public beta for macOS.
What's the catch
Where a developer used to chat with one model at a time, Xirp lets you coordinate a whole army of AI assistants from a single window:
- **50+ parallel sessions.** You can spin up dozens of agents at once: one hunts for bugs, another writes tests, a third updates the docs.
- **Isolated sandboxes.** Agents work in separate workspaces (git worktrees), so they don't step on each other or break the shared codebase.
- **Full freedom (vendor-agnostic).** You can switch between models from different providers (Claude, Gemini, OpenAI) on the fly, within the same task.
- **Deep context.** Thanks to integration with Spotify's internal ecosystem, the AI doesn't just see an isolated chunk of code — it understands the company's architecture and standards as a whole.
The project turns the developer from an individual contributor into a team lead managing a workforce of digital coders.
Why it matters
Xirp joins a broader race to build tools for managing AI coding agents — see also the OpenAI Codex vs. Claude Code rivalry and Meta's always-on Manus agents. Spotify shows that orchestrating dozens of agents at once isn't hypothetical — it's already a working process inside a major company.
You can check out the official announcement and details on Spotify Engineering's X account.