Anthropic ships Claude Mods for customizing Claude Code
Claude Mods let developers customize Claude Code's execution loop and interface using hooks and plugins, Anthropic's Thariq Shihipar said in an interview with Latent Space.
Anthropic is rolling out Claude Mods, a system for customizing the Claude Code harness through hooks and plugins. Mods can rewrite both the harness's execution logic and its interface, Anthropic's Thariq Shihipar said in an interview with Latent Space. Mods can also spawn forked sub-agents that preserve prompt caching, Shihipar said, which keeps the added agents cheaper to run.
Developers are already using Mods to build assumption trackers and model routers that pick which Claude version handles a task, Shihipar said. Others have built interactive dashboards rendered inside Claude Code. Multiple mods can be combined for layered functionality, according to Shihipar, though Anthropic has not published a catalog of what people have built.
Mods follow Claude Projects, which Anthropic shipped earlier. Shihipar described Projects as splitting a single persistent conversation across parallel cloud sessions, similar to how Claude Tag works inside Slack. Shihipar said Claude Tag alone now accounts for 80% of Anthropic's internal cloud usage, though the company has not published that figure separately.
No outlet other than Latent Space has published this interview's details. Anthropic has not said when Mods will reach all Claude Code users or how the feature will be priced. Shihipar said a plugin portal and expanded Artifacts, which now support persistent data storage, shipped alongside Mods.