Browser Search APINo API Key, No Billing

Give Claude Code, Codex, Cursor, or any shell-capable agent structured web search
through the browser on your machine. No API key. No metered search bill.

Built with Rust Stable JSON Authenticated browser state
Claude CodeCodexCursor
~/repos/Local-Search
Claude CodeOpus 4.8 · Claude Max

Kevin · local-search playground · ~/repos/Local-Search

Search for “rust browser automation libraries” and return three results.

⏵⏵ auto mode on · ← for agents

Captured July 21, 2026 with the managed local Chrome profile. Results can change with the search engine.

$0per local search
4search engines
54.3 KiBcrates.io package
0API keys

Ask. Search.Return JSON.

lsearch bridges a coding agent and the browser already on your machine. One shell call goes in; compact search context comes back.

One native command

Search from any shell-capable agent.

No SDK, API key, hosted search account, or agent-specific integration.

Agent

Run one command

Any shell-capable agent runs lsearch.

query + limit
Browser

Search locally

Managed Chrome searches with your local state.

Runs locally
Agent context

Return clean JSON

Only stable result fields reach the agent.

Compact + valid
Returned agent context

Stable fields, not search-page chrome.

The browser stays local. Your agent receives only the result data it can act on.

ranktitleurldomainsnippetcontent?
returned to the agentstdout · application/json
{
    "engine": "duckduckgo",
    "ok": true,
    "query": "rust browser automation libraries",
    "search": {
        "blocked": false,
        "results": [
            {
                "rank": 1,
                "title": "browser_automation — Rust web dev library",
                "domain": "lib.rs",
                "url": "https://lib.rs/crates/browser_automation",
                "snippet": "A modular Rust browser automation library…"
            }
        ]
    }
}

Searches, tokens.Latency and cost.

Two search benchmarks cover 72 local searches and 24 matched requests per provider. A separate July 27 release audit measured a 1.06 MB arm64 macOS binary and 36% faster warm CLI startup.

Benchmark 01

Local JSON vs. rendered search-page snapshots

36 searches per depth · 12 queries × 3 engines

Requested depth fulfilled

Every cross-engine search returned the requested number of results.

72/72
Schema-valid stability runs

58/60 repeated runs also kept the exact same top-three URLs.

60/60
Content pages extracted

Every page returned the full configured 1,200-character text cap.

72/72
local-search JSONinteractive page snapshotVisible command + stdout · o200k_base tokens
3 results36/36 usable searches
Page snapshot8,760.5
local-search309
96.5%less context
10 results36/36 usable searches
Page snapshot8,708
local-search891
89.8%less context
Benchmark 02

local-search vs. hosted search APIs

12 identical queries × two depths · 24 requests per provider

Depth fulfilled24/24
Normalized size53.4 tokens / result
Median latency148.7 ms
API usage$0
ProviderDepth fulfilledRequest + response tokensTokens / resultLatency24-request usage
Exa24/244,472.5881.2501.9 ms$0.324
Brave Search24/2413,104108.6322.0 ms$0.120
Tavily17/241,163259.51,184.4 ms24 credits · $0.192 PAYG
Firecrawl24/2450974.81,520.8 ms48 credits · ≈$0.154 Hobby equivalent

Search source build, July 21, 2026. Benchmark 01 compares visible command text plus stdout with compact interactive snapshots. Benchmark 02 counts serialized requests plus raw responses; all token counts use o200k_base. local-search measured 384.5 ms cold and 6.5 ms from its five-minute local cache. The July 27 native audit used Rust 1.97.1 on arm64 macOS (aarch64-apple-darwin); startup is process launch time, not search latency. Full methodology and runner ↗ · Machine-readable data ↗

Frequently asked questions.

Quick answers about installation, compatible agents, search engines, privacy, and how local-search compares.

01

What is local-search?

local-search is an open-source Rust CLI that turns a local Chrome or Chromium browser into a structured search interface for AI coding agents. The lsearch command searches the web and returns stable JSON containing ranked titles, URLs, domains, snippets, and optional page content.

02

Does local-search need a search API key?

No. local-search uses the browser already running on your machine, so there is no hosted search account, metered search plan, or paid API key. Install it from crates.io and run searches from the shell.

03

Which coding agents can use local-search?

Claude Code, OpenAI Codex, Cursor, OpenClaw, and any other shell-capable coding agent can use local-search. The integration is one CLI command rather than an agent-specific SDK, and the result is returned through standard output as JSON.

04

Which search engines does local-search support?

local-search supports Google, Bing, DuckDuckGo, and Brave Search. Searches run through a managed local browser profile, so results can reflect the region and signed-in browser state that you control.

05

How does local-search compare with hosted search APIs?

In the July 21, 2026 matched-provider benchmark, local-search fulfilled 24 of 24 requested result depths, used 53.4 normalized tokens per result, recorded 148.7 milliseconds median latency, and consumed $0 in hosted API credits. The full benchmark runner and methodology are public in the repository.

06

Does local-search send browser credentials to a hosted service?

local-search does not require a local-search cloud service or hosted account. It controls a browser profile on your machine, sends each query to the selected public search engine, and returns the structured result to the calling agent through standard output.

07

How small and fast is the local-search Rust binary?

The compressed local-search 0.1.2 crates.io package is 54.3 KiB. In the July 27, 2026 arm64 macOS release audit, the compiled lsearch executable measured 1,055,072 bytes (1.06 MB), down from 4,623,984 bytes. Median warm CLI process startup fell from 7.20 to 4.61 milliseconds across 200 alternating launches. These figures describe that machine and build, not search latency.

08

Is local-search free and open source?

Yes. local-search is MIT-licensed open-source software published on GitHub and crates.io. It does not require a local-search subscription, hosted search account, or paid API dependency. Queries still use the selected public search engine through the browser on your machine.

Give agents the web.Skip search bills.