Do I need the 800 MB Laya download for this to work?

No. The base pipeline runs with DuckDuckGo and a lexical heuristic verifier. The Laya download adds neural relevance filtering and stance scoring, and every stage falls back gracefully when it is absent.

How is this different from calling a search API directly?

It inserts a sieve between search and the agent: cross-provider search, Laya or lexical filtering, then stance classification with citation-ready evidence reports instead of raw snippets.

Can I trust the verify verdicts for production claims?

Lexical mode is keyword overlap and is blind to paraphrase and negation scope. Laya mode replaces cutoffs with calibrated probabilities, but the defaults are uncalibrated and still fallible.

What setup catches should I expect on GPUs or locked networks?

It runs on CUDA, Apple Silicon MPS, or CPU with slower throughput, and supports an alternate Hugging Face endpoint. Bright Data and Web Unlocker paths need their own provider setup.

What are the search provider paths?

Free DuckDuckGo by default, self-hosted SearXNG for dual-provider cross-validation, Bright Data for production SERP across seven engines with geo-targeting, and a Web Unlocker path that auto-retries 403, 429, CAPTCHA or geo-blocked pages.

How does Laya's verification blend compare with the lexical one?

In Laya mode the score is 0.55 times neural stance plus 0.35 times source quality plus 0.10 times freshness, against the lexical model's 0.45 times keyword overlap. The result payload keeps the full lexical sidecar so you can compare both classifiers on your own data.

Which Laya checkpoints can I use?

The root default is a ModernBERT-large 421M model for English; multilingual is an mmBERT-base 322M model covering 100+ languages and about twice as fast; typed-decisions is another ModernBERT-large 421M tuned for workflow decisions. Pick one with --laya-subfolder.

CommunityJev Agents & MCP

Zero API Key Web Search

Python CLI and MCP server for keyless web search over DuckDuckGo or self-hosted SearXNG, with an optional local Laya sieve that filters, reranks and stance-checks evidence before it reaches an agent.

See highlights

Open github.com

Stars: 18
Language: Python
License: MIT
Published by: Community
Author: wd041216-bit
Last verified: Oct 2026

wd041216-bit/zero-api-key-web-search

User repository on GitHub

View on GitHub

Jev-powered search infrastructure for AI agents: zero API keys, MCP-ready, LLM-context aware, with local neural evidence verification.

Forks: 4
Last pushed: Updated Sep 2026

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

About this project

Zero-API-Key Web Search ⚡ Jev-Powered ⚡ — free neural search & evidence verification for AI agents. Zero API keys. MCP-ready. Batch-filter, rerank and verify evidence locally at 100+ decisions/sec with an open Jev-class decision model. An agent that searches the raw web pays for everything it reads: context tokens, latency, and attention spread over SEO noise. Zero-API-Key Web Search puts a sieve between search and the agent:

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import

Works forJev for Classification Jev for Tool calling

Used inJev for Software & DevTools

Languages: Python

Highlights

Keyless DuckDuckGo by default, self-hosted SearXNG for cross-validation, Bright Data for production SERP.
Optional local Laya sieve scores relevance and support or conflict stance in one forward pass per item.
MCP server with eight tools; search, context, verify and report calls accept per-call Laya opt-in.
Circuit breaker per provider, response cache, async fan-out, claim decomposition and deep page-aware checks.

Watch out

The lexical verifier is pattern matching and misses paraphrase and negation scope; Laya defaults are uncalibrated so sweep the threshold on your own labeled data. Verification still depends on what search returns. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jev-recruiter

Local LinkedIn recruiting workspace on Browser Use's Jev Ultrafast: Jev screens titles, drives navigation, and assesses profile excerpts against your brief while a live browser stays visible; matches and evidence export as JSON.