Search built for agents,
not for humans.
KlawSearch is the retrieval layer for LLM apps. One API call returns a ranked, cleaned, embedding-aware result set — plus a synthesized answer your agent can ground on. Drop-in compatible with LangChain, Goose, and other agent toolchains.
# one POST. ranked results + synthesized answer. curl https://api.klawsearch.com/v1/search \ -H "Authorization: Bearer ks-live-..." \ -H "Content-Type: application/json" \ -d '{ "query": "latest LLM agent benchmarks", "max_results": 5, "search_depth": "advanced", "include_answer": true, "include_domains": ["arxiv.org", "github.com"] }' ⤷ 200 OK · 1.43s { "answer": "SWE-bench Verified now leads at 67%...", "results": [ { "title": "AgentBench v2.1", "score": 0.94, "url": "arxiv.org/abs/2410.0921" }, ... ], "response_time": 1.43 }
Built for agents, not humans.
Hybrid retrieval with a finance-aware corpus, exposed behind the standard LangChain web-search interface so swapping providers takes one import line.
Hybrid retrieval
BM25 + dense vectors + reranker. Returns markdown + citations, not blue links.
Finance-native
SEC EDGAR filings, earnings, insider trades indexed by ticker. /search/filings out of the box.
Live web fallback
If our curated corpus doesn't have it, our self-hosted meta-search reaches the open web automatically.
Agent-ready
Standard LangChain web-search interface. Drop-in for tools that expect the canonical url/title/content/score shape.
The retrieval primitive for
every agent stack.
From open-source ReAct loops to production trading bots — KlawSearch returns the same well-formed payload, so the rest of your code never has to know.
LangChain
Drop into any agent loop as a BaseTool. Compatible with the standard LangChain web-search tool signature — change the import line and ship.
tool = KlawSearchResults(k=5)
Goose
One YAML entry registers search and extract as native Goose tools. Works locally over stdio or hosted over HTTP.
klawsearch:
command: npx -y @klawsearch/mcp
OpenClaw
First-party integration. Combine generic web search with the Finance API to drive on-chain & SEC-aware trading agents.
agent.use("klawsearch.finance")
Ask anything. See the full response shape your agent gets back.
This is the actual POST /v1/search handler running against a sample corpus. The same JSON your LangChain tool would parse.
{
// run a query →
// answer + ranked results appear here
//
// every response includes:
// answer synthesized, source-grounded
// results[] {title, url, content, score}
// response_time ms latency, end-to-end
}Search APIs, side by side.
Other agent-search APIs nailed the basics, search category — KlawSearch keeps what works, opens what's closed, and pushes into financial data where the incumbents don't yet play.
| KlawSearch | Tavily | Exa | Brave | Serper | |
|---|---|---|---|---|---|
| Hybrid retrievalBM25 + vector + RRF | ✓ | ✓ | ✓ | — | — |
| Cross-encoder rerankbge-reranker / cohere | ✓ | opt | ✓ | — | — |
| Synthesized answergrounded, source-cited | ✓ | ✓ | opt | — | — |
| SEC / financial filingsEDGAR, 10-K, 10-Q, S-1 | ✓ | — | — | — | — |
| Congressional disclosuresPelosi tracker style | ✓ | — | — | — | — |
| Self-hostablesingle Docker compose | ✓ | — | — | — | — |
| Drop-in LangChain toolsame response shape | ✓ | native | — | — | — |
| Free tierno card required | 1,000 / mo | 1,000 / mo | 1,000 / mo | — | 2,500 once |
KlawSearch Finance.
The agent-grade feed for public markets.
SEC filings, congressional trades, earnings transcripts, FOMC minutes, on-chain — normalized, embedded, and queryable through the same API your agents already speak.
Every filing, embedded.
10-K · 10-Q · 8-K · S-1 · proxy statements. Indexed within 60s of publication, queryable by ticker, form type, and exhibit.
Disclosures, not headlines.
Periodic Transaction Reports parsed and joined with ticker metadata. Webhook alerts when a member trades.
Transcripts & FOMC.
Earnings calls and Fed minutes — chunked, embedded, semantically searchable across quarters.
Cross-source signal.
"How did X filing move Yticker?" — one call joins news, filings, on-chain, and price action.
Find a plan to power your AI agents.
One credit = one /search or /extract call. Answer synthesis included. Cached results within 24h are free.
- 1,000 API credits / month
- No credit card required
- LangChain-compatible SDK
- Community Discord
- Pay only for what you use
- Cancel anytime
- Finance API à la carte
- Email support
- 3,500 API credits / month
- LangChain-compatible SDK
- Community Discord
- Email support
- 9,000 API credits / month
- Finance API · SEC + earnings
- Higher rate limits · < 1s p95
- Priority email support
- Custom API call volume
- Custom rate limits
- SLAs · SOC2 audit logs
- On-prem · SSO · BYO LLM
From zero to first call in 60 seconds.
The quickstart drops a working agent + KlawSearch tool into your repo. Inspect the response shape, swap models, point at a self-hosted corpus — all without leaving your editor.
$ export KLAWSEARCH_API_KEY=ks-live-...
>>> from klawsearch import KlawSearch
>>> client = KlawSearch()
>>> r = client.search("AAPL 10-K risk factors")
>>> r.answer
"Apple's FY2025 10-K cites supply-chain concentration in
China, FX volatility, and regulatory pressure on the App
Store as material risks..."
>>> [r.url for r in r.results[:3]]
['sec.gov/.../aapl-10k-2025.htm',
'sec.gov/.../aapl-10q-q3.htm',
'investor.apple.com/sec-filings/...']
Stop scraping. Start retrieving.
One thousand free search credits, every month. No credit card, no demo call, no waitlist for the generic API.