Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

webctl

Web search CLI for agents that has Jev score results and scraped chunks against the query and goal, passing only relevant ones on.

Benchmarkbeta
View on GitHub ↗

Overview

webctl is a search command-line tool for coding agents that aims to save context-window tokens. By default it queries up to three web search backends; each result set, with the original query and an explicit --goal, is passed to Jev for scoring, and the high-scoring subset is deduplicated before reaching the agent.

When snippets are not enough, --scrape --filter-chunks fetches the top pages, splits them into roughly 2,000-character chunks with overlap, and asks Jev which chunks are relevant, so only those land in the agent's context. The author says this saves a great deal on long PDFs, comment threads and full articles but adds nothing for short facts. Benchmarks run Claude Code, Codex and pi on 30 research questions with and without webctl, graded by another model; per the list, one arm that could not be observed was excluded. It installs via Homebrew, apt, npm or go install.

What the decision model decides

TypeQuestionAnswers
ScoreHow relevant is this search result to the query and goal?—
Yes / noIs this chunk of the page relevant to the goal?yes · no

How it works

  1. Run webctl with a query and --goal.
  2. Query up to three search backends.
  3. Jev scores results; keep and dedupe the relevant ones.
  4. Optionally scrape top pages and keep only relevant chunks.

Strengths

  • Less irrelevant text in the agent's context.
  • Chunk filtering for long pages.
  • Works with Claude Code, Codex and pi.

Caveats

  • Benchmark graded by a model and excludes one arm.
  • Adds latency and cost for short factual lookups.

Getting started

webctl setup (enter Jev key), then webctl search "<query>" --goal "<goal>"

At a glance

Type
Benchmark
Status
beta
Language
Go
Runtime
CLI
Has code
yes

Use cases and integrations

How curated lists describe it

  • Search CLI for agents: results from up to three backends are scored by Jev against your query and an explicit `--goal`, and only the relevant ones reach the agent's context. Its benchmark excludes an arm it could not observe.

    MrJev/awesome-jev

Frequently asked questions

What does Jev decide?

Which search results and page chunks are relevant to your query and goal.

When is scraping worth it?

For long pages like PDFs, threads and full articles; not for short facts.

Which agents use it?

Claude Code, Codex and pi, among others.

Related projects

  • Jev Deep Research

    Research agent where GPT plans and Jev finds evidence in parallel across document regions with Choice and Noul.

    Browser extensionresearchTypeScript
  • Jev Search★ 534

    Plain-language web search where a decision model like Jev picks query, sources and time range, then ranks Search1API results.

    AppproductionTypeScript
  • hippo-memory

    Local long-term memory for coding agents with an optional Jev reranker that orders recalled memories.

    ProjectproductionTypeScript
  • sift

    Chrome extension that re-ranks Google results with Jev questions per result and folds away sales pages and SEO filler.

    Browser extensionbetaTypeScript
  • LinkScout

    Browser extension that has Jev score search results for relevance, depth, SEO filler and page type, then re-ranks them with a 0-100 badge.

    Browser extensionprototypeJavaScript
  • jevsearch

    shadcn/ui site-search block that shows keyword hits instantly, then has Jev re-rank the top 20 by what the visitor meant.

    DemobetaTypeScript