Skip to main content
plugin CX Codex

caliber-jev-compaction — Claude Code plugin: what it does

caliber-jev-compaction compacts Claude Code sessions using TypeSafe's Jev model to score and drop stale tool calls instead of summarizing.

MIT ★ 1,319 updated 2026-09-24

Install into Claude Code · Plugin marketplace All methods and platforms ↓
/plugin marketplace add caliber-ai-org/ai-setup
/plugin install caliber-jev-compaction@caliber
Open on GitHub

What caliber-jev-compaction does

The plugin evaluates tool calls with TypeSafe's Jev model and discards obsolete actions rather than generating summaries. This preserves exact tool outputs in the context window while removing dead weight from the history.

Setup requires access to the caliber-ai-org/ai-setup repository. Without adding the marketplace source first, the installation command will fail.

What it helps with

  • compact long Claude Code sessions
  • drop stale tool calls from context
  • keep exact tool outputs without summarization

Example prompts for the agent

  • Run compaction on the current session with caliber-jev-compaction
  • Score tool calls using TypeSafe's Jev model
  • Remove obsolete tool history from the session

How to install caliber-jev-compaction

Plugin marketplace

/plugin marketplace add caliber-ai-org/ai-setup
/plugin install caliber-jev-compaction@caliber

Step by step, with commands for each platform: Claude Code plugins: marketplaces, installing, and building your own

Questions about caliber-jev-compaction

How to install caliber-jev-compaction in Claude Code?

Run two commands in order: /plugin marketplace add caliber-ai-org/ai-setup followed by /plugin install caliber-jev-compaction@caliber.

What does caliber-jev-compaction do?

It compacts Claude Code sessions by scoring tool calls with TypeSafe's Jev model and dropping stale entries instead of summarizing them.

Which model powers the tool call evaluation?

The plugin uses TypeSafe's Jev model to score tool calls during session compaction.

Source text

caliber-jev-compaction · MIT
Verbatim Jev-guided compaction for Claude Code sessions — scores tool calls with TypeSafe's Jev model and drops stale ones instead of summarizing.

© caliber-ai-org · MIT · caliber-ai-org/ai-setup

Similar plugins

All plugins for Agent setup: memory, context, orchestration →

From the same set17

The whole set →
find-skills find-skills discovers community skills from the Caliber registry and installs them directly into your coding session. CX Codex +1 ★ 1,319repo stars save-learning save-learning turns a user's lasting preferences into project learnings. Use it when an agent should carry a rule from one session into the next. CX Codex +1 ★ 1,319repo stars setup-caliber setup-caliber walks a project through Caliber setup: installs the CLI, hooks agent config sync into every commit, and creates the config files your agents are missing. CX Codex +1 ★ 1,319repo stars adding-a-command adding-a-command generates Commander.js CLI commands in src/commands/, connects them to src/cli.ts, and sets up telemetry tracking. CX Codex +1 ★ 1,319repo stars llm-provider The llm-provider skill guides you through adding an LLM backend to a project, from its provider class to configuration and factory registration. Use it when connecting a third-party API or another model. CX Codex +1 ★ 1,319repo stars scoring-checks Adds deterministic scoring checks that measure AI agent config quality. Evaluates file structure, standards compliance, and configuration patterns — with no randomness or external calls. CX Codex +1 ★ 1,319repo stars caliber-testing caliber-testing is a Claude Code skill that generates Vitest tests with process.env resets, vi.hoisted mocks, and test files placed in __tests__ directories. CX Codex +1 ★ 1,319repo stars writers-pattern The writers-pattern skill guides you through adding a module that writes agent config files for a new platform. Use it when extending caliber to another coding tool. CX Codex +1 ★ 1,319repo stars

Project sponsors

Companies that keep this analytics open