Cache Hit plugin for Herdr

Keep the warm context warm.

cache-hit gives Herdr a real-time view of prompt-cache health: hit ratios, retained tokens, expiration countdowns, and deadline-aware agent sorting.

Install the plugin View on GitHub

Works locally with Codex CLI, Claude Code, OpenCode, and Antigravity CLI on macOS and Linux.

Features

One compact sidebar answers the question that matters before you switch agents: which context is still reusable, and which one needs attention now?

Alarm icon shows when cache is about to expire

When a cache is about to expire, the custom expiry sort view pushes its agent to the top. A configurable alarm icon appears, and the countdown becomes bolder.

Active caches show remaining time and retained tokens

Active agents show time to expiry, cache-hit percentage when available, and retained read tokens. The same indicators work across Antigravity CLI, Codex, and Claude.

Cold caches keep token history, not stale percentages

After expiry, the percentage disappears. The ❄ ⇣tokens state keeps retained-token history without implying that the cache is reusable.

A Herdr terminal window with an expiry-sorted sidebar containing Antigravity CLI, Codex, and Claude agents. The top Antigravity CLI row has an alarm countdown; active Codex and Claude rows show cache metrics; cold Claude rows show snowflakes and retained-token counts.

Close-up of the expiry-sorted Herdr sidebar. The Antigravity CLI row at the top shows a red alarm countdown and 154.8 thousand retained tokens.

Alarm icon shows when cache is about to expire

The expiry sort view pushes the agent to the top. A configurable alarm icon appears, and the countdown becomes bolder.

Close-up of active Codex and Claude rows in Herdr, showing healthy expiry countdowns, a cache-hit percentage, and retained-token metrics.

Active caches show remaining time and retained tokens

Active agents show time to expiry, cache-hit percentage when available, and retained read tokens.

Close-up of cold Claude rows in Herdr, each marked with a snowflake and a retained-token count but no stale cache-hit percentage.

Cold caches keep token history, not stale percentages

After expiry, the percentage disappears. The ❄ ⇣tokens state keeps retained-token history without implying that the cache is reusable.

See active cache metrics

See time to expiry, cache-hit percentage when available, and retained read tokens. The values update while the cache is active.

Expired caches show as cold

When a cache expires, its percentage disappears. A snowflake and the retained-token count remain.

Put expiring caches first

Press prefix+s to put the cache closest to expiry at the top of the agents pane. Press it again to return to the default view.

Compatibility: macOS and Linux; Windows through WSL2. jq and Bash are required. Python is needed only for OpenCode monitoring or view toggling; Go is needed only to build the optional Antigravity CLI helper from source.

Install

Terminal
$ herdr plugin install e-kotov/herdr-cache-hit

User guide · GitHub repository · Releases · MIT license

cache-hit is an open-source plugin for Herdr by Egor Kotov.