SearchLane

Agent web search & research API — structured hits, news search, and multi-source research briefs with citations.

Part of Talocode — open tools people trust; hosted power under docs.talocode.site.

Why SearchLane?

Agents and apps need search as a product, not a one-off scrape:

• Structured JSON — title, URL, snippet, score, provider (easy for tools / MCP)
• Three actions — web query, news, multi-source research with citations
• Provider cascade — Brave / Serper when keys exist; free DuckDuckGo path otherwise
• Credit metering — predictable cost on Talocode Cloud (5 / 8 / 30 credits)
• Same contract — npm SDK, Python client, CLI, and HTTP share one API surface

SearchLane is not a full browser (see Agent Browser) and research v0.1 is extractive (ranked sources + brief), not a giant offline LLM.

Install

pip install talocode-searchlane
# or upgrade
pip install -U talocode-searchlane

Requires Python 3.10+. Stdlib only (no extra runtime deps).

Quickstart

import os
from searchlane import SearchLaneClient

client = SearchLaneClient(
    api_key=os.environ["TALOCODE_API_KEY"],
    # base_url defaults to https://api.talocode.site
)

# Structured web search (5 credits)
hits = client.query(query="MCP agent tools", limit=5)
for r in hits.get("results", []):
    print(r["position"], r["title"], r["url"])

# News-biased search (8 credits)
news = client.news(query="AI agents", limit=5)

# Research brief + citations (30 credits)
brief = client.research(query="What is llms.txt?", fetch_pages=True)
print(brief.get("answer"))
for c in brief.get("citations", []):
    print(c["index"], c["title"], c["url"])

Environment

export TALOCODE_API_KEY=your_key
export TALOCODE_BASE_URL=https://api.talocode.site

If base_url already ends with /searchlane (edge-style hosts), paths like /v1/searchlane/query are rewritten to /query automatically.

CLI

After install, the searchlane console script is available:

export TALOCODE_API_KEY=...

searchlane health
searchlane pricing
searchlane capabilities
searchlane query --query "open source AI tools" --limit 5
searchlane news --query "AI agents"
searchlane research --query "agent-native APIs" --fetch-pages

Or: python -m searchlane query -q "...".

API routes & credits

Auth: Authorization: Bearer $TALOCODE_API_KEY or X-Api-Key.

Example response (query)

{
  "query": "MCP agent tools",
  "results": [
    {
      "title": "...",
      "url": "https://...",
      "snippet": "...",
      "score": 0.92,
      "source": "duckduckgo-html",
      "position": 1
    }
  ],
  "provider": "duckduckgo-html",
  "count": 5,
  "durationMs": 900,
  "engine": "searchlane",
  "version": "0.1.0",
  "usage": { "credits": 5, "action": "searchlane.query" }
}

Research responses also include answer and citations[].

cURL

curl -s https://api.talocode.site/v1/searchlane/health

curl -s -X POST https://api.talocode.site/v1/searchlane/query \
  -H "Authorization: Bearer $TALOCODE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query":"MCP agent tools","limit":5}'

How search works (hosted)

Provider chain (automatic):

1. Brave — when the host sets BRAVE_API_KEY
2. Serper — when the host sets SERPER_API_KEY
3. DuckDuckGo — Instant Answer + HTML (no key)
4. mock — offline / fallback demo hits

Safety: public http/https only; localhost and private ranges blocked. Research may peek a few result pages for richer snippets (optional fetch_pages=False to skip).

Client API reference

from searchlane import SearchLaneClient, SearchLaneError, create_searchlane_client

client = SearchLaneClient(api_key=None, base_url=None, timeout=60.0)

client.health()           # dict
client.pricing()          # dict
client.capabilities()     # dict
client.query(query=..., limit=None)
client.news(query=..., limit=None)
client.research(query=..., limit=None, fetch_pages=None)

• SearchLaneError — raised on HTTP failures (status, code when present)
• create_searchlane_client(...) — factory alias

Node / full product

Same cloud product from JavaScript:

npm install @talocode/searchlane

import { SearchLaneClient } from '@talocode/searchlane'
const search = new SearchLaneClient({ apiKey: process.env.TALOCODE_API_KEY })
await search.query({ query: 'MCP agent tools', limit: 5 })

CLI: npx searchlane query --query "...".

Repo (server, SEA binaries, release assets): github.com/talocode/searchlane

Talocode ecosystem

Open-source workflow layers for builders. Sibling projects:

More: github.com/talocode · docs.talocode.site · talocode.site

Links

• PyPI: https://pypi.org/project/talocode-searchlane/
• npm: https://www.npmjs.com/package/@talocode/searchlane
• Source: https://github.com/talocode/searchlane
• Docs: https://docs.talocode.site
• Issues: https://github.com/talocode/searchlane/issues

License

MIT © Talocode