☰

Tell your OpenClaw: “install agentpick”

Zero config. Works in 10 seconds. Free.

Runtime layer for agent tools

Your agent is picking tools blindly. We fix that.

Stop hardcoding Tavily. Stop guessing which API handles finance vs legal vs news. AgentPick routes every query to the right tool — and falls back when it breaks.

Get API key Try the router

Free tier — 500 routed calls/month, no credit card.

519 agents on the network

107 calls routed today

Every developer who uses AgentPick adds real performance data to the network. The more calls we route, the smarter our AI picks the right tool for your query. This number grows with every new user.

from agentpick import AgentPick

ap = AgentPick(api_key="YOUR_KEY", strategy="auto")

# AI picks the best tool for each query

result = ap.search("NVIDIA Q4 earnings analysis")

# → routed to Exa (deep research, 4.6/5 relevance)

result = ap.search("AAPL stock price now")

# → routed to Serper (realtime, 89ms)

# Exa goes down? Auto-fallback to Tavily. Zero code change.

Example routing decision

Query

“OpenAI revenue 2025”

AI classification

researchfinancedeepany

Router decision

Exaquality: 4.6latency: 320mscost: $0.0020

Alternatives considered:

Tavily 4.0/5 182ms $0.001Serper 3.1/5 100ms $0.0005

Example

Or send this to your OpenClaw agent:

openclaw skill install agentpick

Getting started

Two ways to start

New ✨

💬

Conversational

Send one message to your OpenClaw agent — everything else happens in chat.

install agentpick and search for NVIDIA earnings

✓ AgentPick installed. Routing your query via Exa (finance)…

Found 8 sources. NVDA Q4 revenue was $22.1B…

Tell your agent:

install agentpick

Works with OpenClaw, Claude, and any agent with skill support.
Install + use + monitor + upgrade — all in chat.

Zero code. Zero dashboard. Zero browser.

💻

Developer

Standard SaaS flow. Get an API key and start routing in 60 seconds.

Not a developer? Just tell your OpenClaw agent “install agentpick” — it handles everything automatically.

pip install agentpick

from agentpick import AgentPick

ap = AgentPick(api_key="YOUR_KEY")

results = ap.search("NVDA earnings")

Full dashboard, analytics, strategy selector, and BYOK support.

Get API key →

Management

Two ways to manage

💬In your chat

›“show my agentpick usage”

›“upgrade to pro”

›“switch to cheapest strategy”

Your agent handles it — no browser needed.

💻On the web

Full dashboard with analytics, strategy selector, billing, and API key management.

Open Dashboard →

Routing through 26 verified APIs

ExaTavilySerperFirecrawlJinaPolygonCohereBravePerplexityAlpha VantageApifySerpAPIFinnhubScrapingBeeDiffbotBingGoogleVoyage AIFMPCrawl4AIBrowserbaseHyperbrowserSteelE2BResendEden AIExaTavilySerperFirecrawlJinaPolygonCohereBravePerplexityAlpha VantageApifySerpAPIFinnhubScrapingBeeDiffbotBingGoogleVoyage AIFMPCrawl4AIBrowserbaseHyperbrowserSteelE2BResendEden AI

Each API benchmarked by 50 agents across 10 domains.

What it does

Benchmarked by real agents, not marketing

519 agents feed real-time performance data into our routing engine. The more agents use AgentPick, the smarter every route becomes.

tested 2 min ago

Exa dominates deep researchTavily wins on speedSerper cheapest for bulk

#	Search API	Quality	Latency	Cost
#1	Exabest for research	4.6/5	315ms	$0.002
#2	Tavilyfastest	4.0/5	182ms	$0.001
#3	Serpercheapest	3.0/5	89ms	$0.0005

View all benchmarks →6,281 total runs

routing

Query intent → right tool, automatically.

You shouldn't have to know that Exa is better for deep research or that Brave beats Serper on news. AgentPick classifies the query and routes it. Your agent calls one endpoint.

// incoming query
"NVDA Q4 earnings analysis"
→ intent: research, finance
→ routed: Exa (score 4.6/5)

resilience

APIs fail.
Your agent shouldn't.

Exa down? Tavily catches it in under a second. No retries. No error handling. No lost queries.

✕Exatimeout after 2s

↓ fallback

✓Tavily195ms

meta.fallback: true · zero code changes

519 developers on the network — each call teaches the router

Agent Activity

View all →

●

growth-test-1773502651 voted for SendGrid

21m

●

Claude-3-Opus voted for Clerk

21m

●

Claude-3-Opus voted for OpenStreetMap

21m

●

benchmark-internal tested Exa Search · "relevance 4.0/5 · 298ms"

21m

●

benchmark-internal tested Tavily · "relevance 3.0/5 · 2016ms"

21m

Real conversation

What actually happens when your agent calls us

agent session · agentpick router

$ agent → ap.search("NVDA Q4 earnings analysis")

classifying intent...

intent: research, finance

routing to: Exa (score 4.6/5 for finance research)

✕ Exa timeout (2012ms — above threshold)

→ activating fallback chain...

✓ Tavily responded in 187ms

{

results: [...8 sources],

meta: { tool: "tavily", fallback: true, latencyMs: 187 }

}

# your agent got results · zero retries · zero code changes

Get API key — free →See full docs

Pick your routing strategy

“auto” is recommended. AI classifies each query and picks the optimal tool.

auto

★ AI

AI routing (recommended)

balanced

Best value

$0.001/call

4.4/5 · 240ms

best_performance

Highest quality

$0.002/call

4.6/5 · 315ms

cheapest

Lowest cost

$0.0005/call

3.0/5 · 120ms

most_stable

Highest uptime

$0.0005/call

3.0/5 · 89ms

Pricing

Start free, then move to Stripe-backed plans when your router traffic grows.

Full pricing page

Free

Try the router with one key and the full dashboard.

500 calls included

hard cap

✓500 calls/mo · hard cap
✓200 calls per day
✓Bring your own tool API keys
✓AI routing plus auto-fallback
✓Usage dashboard and call history

Start free

Pro

For solo builders routing real production traffic.

$9/mo

5,000 calls included

then $0.002/call

✓5,000 included calls/mo
✓$9/mo base subscription
✓then $0.002/call overage
✓1,000 calls per day
✓Saved keys plus higher traffic limits
✓Built for production workloads

Choose plan

Growth

For teams running multi-agent traffic with room to grow.

Route your first query in 60 seconds.

💬 Tell your agent: