P402

io.github.Z333QPublished by GitHub account Z333Qv1.0.0Updated Nov 10, 2026

Multi-provider AI router with x402 micropayment settlement on Base L2. 1% flat fee.

VerifiedSTDIODesktop onlyFinanceAI & ML

Overview

AI-generated overview

P402 is a local MCP server that routes assistant requests across 300+ LLM providers and settles each request in USDC on Base via x402 micropayments.

What you can do
  • Send chat requests through one OpenAI-compatible endpoint that picks a provider by cost, quality, speed or balanced mode
  • Create and inspect session budgets that cap what an autonomous agent can spend
  • Settle per-request payments on-chain through the x402 facilitator, with gasless USDC transfers on Base
  • Exchange structured tasks with other agents over the A2A JSON-RPC protocol
  • Review real-time spend analytics and cost optimization suggestions
When to use it
Use it when you want an assistant to call many LLM providers through a single key, to steer requests toward cheaper or faster models, or to give an autonomous agent a hard spending cap with on-chain settlement.
Requirements
Runs locally over stdio, typically via npx with the @p402/mcp-server package, so Node.js is needed. Requires a P402 account and an API key supplied in the P402_API_KEY environment variable. Network access to the P402 router is required; the base URL can be overridden with P402_ROUTER_URL. A VS Code extension bundles the same server.
Before you install
Requests are billed: a 1% flat fee applies and each call is settled in USDC on Base, so real funds are spent. The API key is a live credential and should be kept out of shared configs. Session budgets are the main guardrail against runaway agent spending, so set them before letting an agent run unattended. Prompts and responses pass through the P402 router and third-party model providers.

Installation

In SourceWeft

  1. Open P402 in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

P402 Protocol

[npm @p402/sdk] [npm @p402/cli] [npm @p402/mcp-server] [npm @p402/mpp-method] [VS Code Marketplace] [License: MIT] [Docs]

AI payment router. Route across 300+ models, settle per request in USDC on Base or USDC.e on Tempo.

P402 sits between your AI application and every LLM provider. It handles intelligent multi-provider routing (cost / quality / speed / balanced), on-chain micropayment settlement via the x402 protocol and mppx on Base and Tempo, and spending guardrails for autonomous agents.


Why P402

ProblemP402 Solution
Hardcoded to one AI providerRoute across 300+ models automatically
$0.30 payment fees kill micropaymentsUSDC on Base: fractions of a cent per settlement
No spending limits for AI agentsSession budgets + AP2 mandate governance
Fragmented provider APIsOne OpenAI-compatible endpoint
No visibility into AI costsReal-time analytics + optimization suggestions

Quick Start (VS Code / Cursor / Windsurf)

Install the extension — the MCP server is embedded, tools appear in Copilot agent mode immediately, no config files required:

ext install p402-protocol.p402

Then run P402: Configure API Key from the command palette.

→ VS Code Marketplace · Open VSX


Quick Start (Claude Desktop / any MCP client)

json
{  "mcpServers": {    "p402": {      "command": "npx",      "args": ["-y", "@p402/mcp-server"],      "env": { "P402_API_KEY": "p402_live_..." }    }  }}

→ MCP docs · MCP Registry


Quick Start (SDK)

bash
npm install @p402/sdk
typescript
import P402Client from '@p402/sdk';
const p402 = new P402Client({ apiKey: process.env.P402_API_KEY });
// Drop-in OpenAI replacement — P402 picks the best providerconst response = await p402.chat({  messages: [{ role: 'user', content: 'Explain x402 payments in one sentence.' }],  p402: { mode: 'cost' }   // cost | quality | speed | balanced});
console.log(response.choices[0].message.content);// p402_metadata: { provider: 'deepseek', cost_usd: 0.00031, latency_ms: 412 }

Quick Start (CLI)

bash
# Authenticate oncenpx p402 login
# Chat using the cheapest providernpx p402 chat "What is x402?" --mode cost
# Check facilitator healthnpx p402 health

Routing Modes

ModeOptimizes ForTypical Provider
costLowest priceDeepSeek V3, Haiku 4.5, GPT-4o-mini
qualityBest outputClaude Opus 4.6, GPT-5, Gemini 3 Pro
speedLowest latencyGroq LPU, Flash models
balancedEqual weight (default)Sonnet 4.6, GPT-4o, Gemini Flash

Session Budgets

Enforce hard spending caps for autonomous agents:

typescript
// Create a $10 session — agent cannot spend a cent moreconst session = await p402.createSession({ budget_usd: 10 });
// All chat requests are deducted from the sessionconst response = await p402.chat({  messages,  p402: { session_id: session.id, mode: 'cost' }});
// Check remaining budgetconst { budget } = await p402.getSession(session.id);console.log(`$${budget.remaining_usd} remaining`);

x402 Payments

x402 is a machine-native payment protocol using HTTP 402. AI agents pay for resources using gasless EIP-3009 USDC transfers on Base L2.

Client → signs EIP-3009 authorization       → POST /api/v1/facilitator/verify       → POST /api/v1/facilitator/settleFacilitator → executes transferWithAuthorization            → pays gas (user pays zero gas)            → returns { success, txHash, receipt }

Network: Base Mainnet (Chain ID: 8453) · Asset: USDC 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913

→ x402 payments guide


A2A Protocol

P402 implements the Google A2A spec over JSON-RPC 2.0. Agents communicate through structured tasks, discover capabilities via /.well-known/agent.json, and settle payments via the x402 extension.

typescript
// Discover P402's capabilitiesGET https://p402.io/.well-known/agent.json
// Submit a taskPOST https://p402.io/api/a2a{ "jsonrpc": "2.0", "method": "tasks/send", "params": { ... } }

→ A2A protocol guide


Packages

PackageDescriptionVersion
@p402/sdkTypeScript SDK — P402Client, types, EIP-712 mandate helpers[npm]
@p402/cliCLI tool — login, chat, sessions, mandates, analytics[npm]
@p402/mcp-serverstdio MCP server — 6 tools over Model Context Protocol[npm]
@p402/mpp-methodmppx payment methods — Base EIP-3009 and Tempo TIP-20 multi-rail settlement[npm]
p402 VS Code extensionEmbedded MCP server for VS Code, Cursor, and Windsurf — zero config[Marketplace]

Examples

ExampleWhat It Shows
01-quickstartLogin → chat → view spend in ~20 lines
02-openai-migrationDrop-in OpenAI SDK replacement
03-nextjs-session-budgetBudget-capped AI in a Next.js App Router project
04-a2a-agentsTwo agents communicating with x402 payment gate

Docs

Guide
Getting StartedAccount, API key, first request
AuthenticationAPI keys, env vars, security
Routing GuideModes, scoring, providers, models
x402 PaymentsEIP-3009, wire format, settlement
SessionsSession lifecycle + budget enforcement
A2A ProtocolJSON-RPC, mandates, Bazaar
CLI ReferenceFull CLI command reference
OpenAPI SpecMachine-readable API schema

Community


License

MIT © P402 Protocol

Source: README.md at commit 94c8cc9

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v1.0.0LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts