What does openclaw-jev-compaction do?

It is a context engine for OpenClaw: Jev scores which tool calls and tool results a session still needs, the rest are dropped, and nothing is summarized. What stays is the original text, word for word, so paths, errors and constraints survive.

How is it different from summarization-based compaction?

Summaries are lossy; a file path, an exact error or a constraint can disappear when the model needs them later. This engine deletes only spent tool calls and results and never rewrites user or assistant text.

What does a compaction pass drop?

Spent tool calls and their results, such as a directory listing already used, a passing test run, or a file read before it was rewritten. User and assistant text is protected, and the log shows how many calls were dropped, truncated, kept, pinned or protected.

Run openclaw plugins install openclaw-jev-compaction --accept-capabilities, which consents to taking OpenClaw's exclusive context-engine slot, then restart the Gateway. Supply the TypeSafe key via the Gateway's TYPESAFE_API_KEY or openclaw.json.

When does compaction run?

At assemble before every model call, at afterTurn when the session crossed the threshold, at compact on /compact and overflow recovery, and at commitTurn to record advancement keys so retried turns stay idempotent. compactAtPercent defaults to 60.

What happens if verbatim compaction cannot fit the budget?

The engine hands the session over to OpenClaw's summarizer as fallback. You can also go back to the built-in behavior entirely by setting contextEngine to legacy or removing it; the README says sessions carry on either way.

Who built the plugin?

The Hilo team, inspired by tamaratran/fast-jev-compaction, the Claude Code plugin that first put the idea to work. It is MIT licensed, published on npm, and developed with CI on GitHub.

Jev directory

CommunityJev Integrations & Adapters

OpenClaw Jev Compaction

A context engine for OpenClaw that asks Jev which tool calls and results a session still needs, drops the rest and never summarizes: what stays is the original text, while one request scores dozens of calls in about a second.

See highlights Open github.com

Stars: 1
Language: TypeScript
License: MIT
Reading time: 2 min
Published by: Community
Author: sqaassl
Last verified: Oct 2026

sqaassl/openclaw-jev-compaction

Organization repository on GitHub

View on GitHub

Verbatim context compaction for OpenClaw: a context engine powered by TypeSafe's Jev. Drops stale tool calls and results, never summarizes.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

Tagscommunity catalog-import logicrw openclaw compaction context-engine

Works forJev for Decision gating Jev for Tool calling

Used inJev for Software & DevTools

Languages: TypeScript
Platforms: CLI
Capabilities: Open sourceRequires API key

Highlights

Jev scores dozens of calls in about a second per request; user and assistant text is never rewritten.
One logged pass reports 71% fewer chars, 9 calls dropped, 3 results truncated, 2 kept, 4 pinned and 5 protected.
Runs at assemble, afterTurn, compact and commitTurn; a forced /compact pass can fall back to OpenClaw's summarizer if verbatim compaction cannot fit.
Install with one command and accept the exclusive context-engine slot; key via TYPESAFE_API_KEY or openclaw.json; compactAtPercent defaults to 60.
Built by the Hilo team and inspired by tamaratran/fast-jev-compaction, the Claude Code plugin that tried this pattern first.

Quickstart

bash

openclaw plugins install openclaw-jev-compaction --accept-capabilities

Watch out

Grounded in the truncated README and GitHub metadata reviewed on 2026-10-15; the README cuts off mid description of a pass, its 71% figure comes from one logged run rather than a benchmark, and taking OpenClaw's exclusive context-engine slot changes compaction for every session.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

X: The compaction plugin, in Chinese

X: The compaction plugin, in Spanish

From the Jev guides

Original write-ups on what this entry is for.

Decision gates in agent frameworks: what each integration ships

Oct 16, 2026

How Pydantic AI, Mastra, TanStack AI, DSPy, Spring AI, and LiteLLM expose Jev: where the typed decision sits in the loop, what each integration returns, and which big frameworks are only adjacent.

Jev Integrations & Adapters#agents#frameworks#integrations

Read article

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai