Agent-Friendly Services

English | 简体中文

Which services let an AI Agent actually get the job done? We have Agents try real tasks and record pass rate, cost and access requirements.

195 services · 35 tested · 187 valid trials · latest 2026-10-10

Pass rate: share of valid trials, across all tasks in the subcategory including setup, that passed independent review (passed / valid trials). Avg. cost: mean per trial of estimated model cost (LiteLLM rates) plus recorded service charges. No signup: ✅ usable without registering first · ❌ an account or key must be prepared first. Runs invalidated by environment faults are excluded. Use → beside a subcategory for per-task results.

Travel · Databases · Search & Web Scraping · Maps & Geography · Weather & Calendar · Financial Data · Scholarly & Reference · Productivity & Collaboration · Developer Tools · Payments / Billing · Communication

Travel

Flights →

Service	Pass rate	Avg. cost	Via	No signup
Kiwi.com	100% (2/2)	$0.85	MCP	✅
Ignav Flights	100% (1/1)	$0.93	Web	✅

Databases

Hosted Relational Databases →

Service	Pass rate	Avg. cost	Via	No signup
Turso	92% (11/12)	$0.02	API	❌
Neon	88% (7/8)	$0.01	API	✅

Search & Web Scraping

Web Search →

Service	Pass rate	Avg. cost	Via	No signup
Exa	100% (5/5)	$0.01	MCP	✅
Firecrawl	100% (3/3)	$0.01	API	✅
Exa	0% (0/1)	—	API	❌

Web Content Extraction →

Service	Pass rate	Avg. cost	Via	No signup
Jina AI	100% (7/7)	$0.0071	API	✅
Firecrawl	83% (10/12)	$0.01	API	✅
Exa	80% (4/5)	$0.02	MCP	✅

Maps & Geography

Geocoding →

Service	Pass rate	Avg. cost	Via	No signup
Nominatim Public API	100% (2/2)	$0.0094	API	✅
Photon Public Demo API	100% (2/2)	$0.01	API	✅

Route Planning →

Service	Pass rate	Avg. cost	Via	No signup
OSRM Public Demo API	100% (2/2)	$0.0093	API	✅
Valhalla Public Demo API	100% (2/2)	$0.01	API	✅

Weather & Calendar

Weather Data →

Service	Pass rate	Avg. cost	Via	No signup
MET Norway Locationforecast	100% (2/2)	$0.0085	API	✅
Open-Meteo	100% (2/2)	$0.0097	API	✅

Public Holidays →

Service	Pass rate	Avg. cost	Via	No signup
OpenHolidays API	100% (8/8)	$0.0096	API	✅
Nager.Date	100% (8/8)	$0.01	API	✅

Financial Data

Exchange Rates →

Service	Pass rate	Avg. cost	Via	No signup
ECB Data Portal API	100% (7/7)	$0.01	API	✅
Frankfurter	100% (4/4)	$0.0056	API	✅

Asset Prices →

Service	Pass rate	Avg. cost	Via	No signup
Alpha Vantage	80% (4/5)	$0.01	API	❌
Alpha Vantage	75% (3/4)	$0.02	MCP	❌

Company Financials →

Service	Pass rate	Avg. cost	Via	No signup
Alpha Vantage	100% (4/4)	$0.01	API	❌
SEC EDGAR Data APIs	100% (2/2)	$0.0092	API	✅
Alpha Vantage	100% (2/2)	$0.03	MCP	❌

Transaction Disclosures →

Service	Pass rate	Avg. cost	Via	No signup
CapitolExposed	90% (9/10)	$0.02	API	✅
Bargo Congress Trades API	55% (6/11)	$0.02	API	✅
Tracefour	0% (0/1)	$0.0071	API	✅

Economic Indicators →

Service	Pass rate	Avg. cost	Via	No signup
ECB Data Portal API	100% (8/8)	$0.02	API	✅
Eurostat Data API	100% (6/6)	$0.02	API	✅

Scholarly & Reference

Scholarly Literature Search →

Service	Pass rate	Avg. cost	Via	No signup
Crossref	100% (2/2)	$0.0097	API	✅
OpenAlex	100% (2/2)	$0.0099	API	✅

Productivity & Collaboration

Collaborative Tables →

Service	Pass rate	Avg. cost	Via	No signup
Grist	100% (7/7)	$0.03	MCP	❌
Notion	100% (2/2)	$0.75	API	❌
Grist	78% (7/9)	$0.03	API	❌

Developer Tools

Dependency Security Advisories →

Service	Pass rate	Avg. cost	Via	No signup
GitHub	100% (8/8)	$0.01	API	✅
OSV.dev	100% (8/8)	$0.01	API	✅

QR Code Images →

Service	Pass rate	Avg. cost	Via	No signup
goQR QR Code API	100% (2/2)	$0.0068	API	✅
QuickChart	100% (2/2)	$0.0083	API	✅

Payments / Billing

Billing · Payment Acceptance →

Service	Pass rate	Avg. cost	Via	No signup
Paddle	0% (0/1)	$1.20	API	❌
paas.build	invalid run	—	API	❌

Communication

Mailboxes →

Service	Pass rate	Avg. cost	Via	No signup
Mail.tm	100% (3/3)	$0.14	API	✅
Guerrilla Mail	50% (1/2)	$0.32	API	✅
AgentMail	0% (0/1)	$0.03	API	✅

How we test

Each trial runs in a fresh session and isolated container. The tested Agent gets only the user task and service entry point, never answers or other services' results; a separate session reviews it against frozen criteria. Usage comes from runner capture and evidence carries hashes, so results can be checked and rerun. Samples are still small: these are observations, not proof of long-term reliability. Method

Latest findings

• Web extraction (GOV.UK holiday table): Jina 3/3, Firecrawl 1/3. Firecrawl fetched the page every time and its HTML was correct, but its Markdown placed year captions after the tables; the Agent twice read the wrong year. Evidence
• Test location changes outcomes. Jina Reader was unreachable from a mainland China server because of DNS interference and was recorded as an invalid run; from an overseas server, setup and both extraction tasks completed. Evidence
• Grist shared expenses: MCP 3/3, REST 2/3. In the REST failure the ledger and formulas were already correct on the service, but the Agent ran out of requests before its final answer: the service worked, the delivery did not. Evidence

Full directory

Every listed service, including untested ones, with each task's text, configuration and per-trial results:

• Travel — 27 services, 2 tested
• Databases — 13 services, 2 tested
• Search & Web Scraping — 12 services, 3 tested
• Maps & Geography — 7 services, 4 tested
• Weather & Calendar — 4 services, 4 tested
• Financial Data — 34 services, 8 tested
• Scholarly & Reference — 3 services, 2 tested
• Productivity & Collaboration — 13 services, 3 tested
• AI Services — 25 services, 0 tested
• Agent Infrastructure & Automation — 5 services, 0 tested
• Developer Tools — 10 services, 4 tested
• Cloud Computing & Hosting — 11 services, 0 tested
• Payments / Billing — 19 services, 1 tested
• Communication — 23 services, 3 tested
• E-commerce — 1 services, 0 tested

Want a service or task tested? Open an issue.

For your Agent

Query guide · Catalog JSON · Results JSON · MCP setup

Methods and contributions

Principles · Inclusion standards · Flight findings · Task design · Execution and review · All results and evidence · Contributing

Code: MIT · Data: CC BY 4.0