Live on Base Mainnet — real USDC payments via x402

Your agent can think.
Now it can act.

One endpoint gives your AI agent 20+ real-world capabilities. Take screenshots. Generate PDFs. Search the web. Validate data. Process documents. No API keys, no accounts — just describe the action and pay in USDC.

request.sh
// One request. Any action.

POST https://api.pylonapi.com/do

{
  "task": "take a screenshot of stripe.com/pricing"
}

// → Returns a PNG screenshot. $0.01. 200ms.

// Or chain multiple actions

POST https://api.pylonapi.com/do/chain

{
  "steps": [
    { "api": "web-scrape", "params": {"url": "..."} },
    { "api": "md-to-pdf", "params": {"markdown": "{{steps.0.output}}"} }
  ]
}
All systems operational
20+ capabilities  ·  100% uptime  ·  Base Mainnet

LLMs can reason. They can't execute.

Your agent can write a business plan, but it can't take a screenshot, generate a PDF, validate an email address, or look up a domain. Every agent builder ends up duct-taping together a dozen APIs with different auth flows, failure modes, and billing systems.

The AI stack has three layers: Brains (OpenAI, Anthropic, Google), Hands (Pylon), and Money (stablecoins via x402). Pylon is the missing middle — the layer that turns reasoning into real-world results.

One request. Any action.

Your agent sends a task in plain English. Pylon matches it to the right capability, executes it, and returns the result. Payment happens automatically via USDC micropayments on Base.

1

Describe the Action

Send a task to /do — natural language or explicit capability. No docs to read, no SDKs required.

2

Pylon Executes

The gateway matches, routes, and executes — with automatic retries, circuit breakers, and quality tracking.

3

Pay & Receive

USDC micropayment on Base via x402. Pennies per action. You get the result in one round-trip.

⛓️

Chain Actions

Use /do/chain for multi-step workflows: "screenshot this page, extract the text, generate a report."

20+ capabilities and growing

Every capability your agent needs to interact with the real world. Screenshots, documents, search, data validation, file processing — all through one endpoint.

📸

See the Web

"Screenshot stripe.com" — Full-page captures in PNG or JPEG. Your agent can see any webpage.

$0.01 / action
📄

Generate Documents

"Convert this markdown to PDF" — Create PDFs from markdown or HTML. Parse existing PDFs. OCR scanned images.

$0.01 / action

Search Anything

"Search for x402 protocol news" — Real-time web search results. No API keys needed from the agent.

$0.003 / action

Validate & Verify

"Validate hello@stripe.com" — Email validation, domain intelligence, WHOIS lookups, DNS records, IP geolocation.

$0.002 / action
🔄

Process Data

"Convert this CSV to JSON" — Format conversion between JSON, CSV, XML, YAML. Resize images. Generate QR codes.

$0.005 / action
🕸️

Extract Content

"Scrape the pricing page" — Extract readable content from any URL. Returns clean markdown with full browser rendering.

$0.01 / action
View all 20+ capabilities →

Two lines of config. 20+ tools.

Install the Pylon MCP server and give Claude, Cursor, or any MCP-compatible assistant instant access to every capability.

// Claude Desktop or Cursor config
{
  "mcpServers": {
    "pylon": {
      "command": "npx",
      "args": ["-y", "@pylonapi/mcp"]
    }
  }
}

That's it. Your assistant now has screenshots, PDFs, search, scraping, and 16 more tools.

npm: @pylonapi/mcp → View Source ↗

Natural language or explicit

Two ways to call any capability — describe what you need, or specify exactly.

Natural Language

{
  "task": "validate the email ceo@stripe.com"
}

Pylon figures out which capability to use and extracts parameters automatically.

Explicit

{
  "capability": "email-validate",
  "params": { "email": "ceo@stripe.com" }
}

Or specify the exact capability and parameters when you know what you want.

Pylon finds what it doesn't have

When an agent requests an action Pylon doesn't natively support, it searches the x402 ecosystem for a service that can fulfill it — and routes the request automatically.

1

No Match Found

Agent asks Pylon to send an email. Pylon doesn't have email natively.

2

Bazaar Search

Pylon searches the x402 bazaar for services that can handle it. Finds an email API at $0.005.

3

Route & Execute

Pylon activates the service, adds reliability tracking, and fulfills the request — one round-trip.

Transparent pricing. Every response shows the full cost breakdown — what the provider charges and what Pylon adds for discovery and routing.

"pricing": {
  "source": "discovered",
  "providerCost": "$0.005",
  "pylonFee": "$0.005",
  "totalCost": "$0.010"
}

Actions should be fast and cheap.

Your agent's reasoning time is expensive. Don't waste it on tasks that should be deterministic API calls.

🧠
Agent does it itself
~15s
reasoning time
$0.30 in LLM costs
Unreliable
Agent calls Pylon
200ms
execution time
$0.01 per action
Guaranteed

Intelligence → Action → Payment

Three layers are emerging in the AI agent economy. Pylon is the action layer.

🧠
Brains
OpenAI, Anthropic, Google — the reasoning layer
Hands — Pylon
The action layer — 20+ capabilities today, unlimited via discovery
You are here
💰
Money
x402, USDC, stablecoins — agents pay per action, no accounts

Built for agents, not dashboards

Traditional APIs require accounts, dashboards, and billing cycles. Pylon requires one thing: payment.

No API Keys

No registration. No OAuth. No tokens to rotate. Just HTTP + USDC.

🛡️

Reliable by Default

Circuit breakers, automatic retries, quality tracking. Every action goes through the reliability layer.

🔌

Open Provider Ecosystem

Anyone can add capabilities. More providers → more actions → smarter agents.

Add your API. Earn per action.

Plug your API into Pylon and every agent in the ecosystem gets access. You keep 85% of revenue. Zero customer acquisition cost.

Become a provider → Read the Guide ↗
20+ capabilities live · Base Mainnet · x402 native

Works with everything you use

Drop Pylon into your existing stack. MCP server, framework SDKs, or raw HTTP — however your agents work.

Agent Frameworks

🦜
LangChain
Python SDK
🚀
CrewAI
Python SDK
Vercel AI SDK
TypeScript
🔌
MCP
Any MCP client

Works with MCP Clients

Claude Desktop Cursor Windsurf OpenClaw Claude Code

Listed On

Built On

🔵 Base 💲 USDC 🔗 x402 Protocol

Give your agent hands.

Install the MCP server, or point your agent at api.pylonapi.com/do. That's the entire integration.

Install MCP Server → View Capabilities ↗