webctl

All projects · Search and retrieval

Smart web-search CLI for agents: query multiple search backends, then use TypeSafe Jev to score, judge, and optionally chunk-score page content so the agent reads less junk.

At a glance	Details
Source	Source
Maintainer	dorkitude. Independently curated; this entry is not an upstream submission or endorsement.
Format	Go CLI (webctl); also installable via Homebrew, apt, npm wrapper, or go install.
Requirements	Go toolchain for from-source builds; live judging uses a TypeSafe/Jev key via webctl setup. Search backends may need their own API keys (Brave preferred).
License	MIT.
Disclosure	AI-assisted catalog review; no affiliation with the maintainer. Listing is not an endorsement. Source and offline go test inspected. Live TypeSafe and live web-search calls were not run for this listing.

When to use

Use it when an agent needs web search results filtered by topic/goal before consuming them, or when optional scrape+chunk scoring should shrink long pages. Prefer a plain search API or jegrep/sgrep when you are searching a local codebase instead of the public web.

How it works

Search providers return candidate results; internal/jev posts typed System One questions (jev-latest prompts under prompts/) for relevance, source quality, dedupe assist, and optional chunk scoring. Application code owns provider selection, deterministic dedupe, scrape parsing, and what the agent sees. Without a Jev key, search can still run against keyless providers with throttling—scoring paths that need Jev will not apply the same filter quality.

Get started

git clone https://github.com/dorkitude/webctl.git
cd webctl
git checkout 903f0f4ddf30cac84ffedc7e4e53dff36fc4846a
go test ./...
# Live path (charges TypeSafe + any keyed search providers):
# go install ./cmd/webctl && webctl setup && webctl search "…" --goal "…"

Upstream also documents Homebrew/apt/npm installs. Do not paste real API keys into the catalog tree.

Examples and demos

• README quick-start and schematics for filter-only vs scrape+chunk flows.
• Offline unit coverage under internal/jev, evals, and cmd/webctl/cli (fake Jev HTTP).

Limits and data handling

Queries, goals, result metadata, and optionally scraped page chunks leave the host on live Jev and search-provider calls. Cost/latency claims in upstream docs are author-reported, not remeasured here. Provider rate limits and keyless cooldowns apply.

Review and maintenance

Reviewed on 2026-09-20 at commit 903f0f4: MIT. AI-assisted source review of the Jev client, prompts, and README. go test ./... passed offline (packages including internal/jev, evals, internal/provider, cmd/webctl/cli). No live TypeSafe or live search calls.

Related: jegrep, sgrep, jev-reranker.