AgentWyre MCP Server

AI ecosystem intelligence for your agent — via Model Context Protocol.

Quick Start

npm install agentwyre-mcp

Add to your MCP config (e.g., Claude Desktop, Cursor, OpenClaw):

{
  "mcpServers": {
    "agentwyre": {
      "command": "npx",
      "args": ["agentwyre-mcp"],
      "env": {
        "AGENTWYRE_API_KEY": "your_key_here"
      }
    }
  }
}

No API key? Free tier gives you 5 signals with a 2-day delay. Your agent can upgrade itself by paying with USDC on Base — no human needed. International users: USDC avoids bank fees entirely.

Tools

| Tool | Description | Tier | |------|-------------|------| | agentwyre_status | Service status + your access level | All | | agentwyre_signals | Latest intelligence signals (curated, hype-checked) | All (limited on free) | | agentwyre_flash | Breaking flash signals | Pro | | agentwyre_search | Search signal archive | Daily / Pro | | agentwyre_security | Security advisories for your dependencies | Pro | | agentwyre_costs | AI provider cost tracking | Pro | | agentwyre_subscribe | Get USDC payment instructions | All | | agentwyre_verify_payment | Verify payment and get API key | All |

Tiers

• Free — 5 signals, 2-day delay. No key required.
• Daily ($2.99/mo) — Full daily feed, same-day, search.
• Pro ($9.99/mo) — Everything + hourly flash, security, costs.

Agent Self-Subscription

Your agent can subscribe itself:

1. Call agentwyre_subscribe with tier "daily" or "pro"
2. Send USDC to the returned wallet address on Base
3. Call agentwyre_verify_payment with the transaction hash
4. Receive API key → set as AGENTWYRE_API_KEY

No credit card. No human approval. Fully autonomous.

What You Get

Every signal is:

• Sourced — Links to original sources
• Hype-checked — Rated: verified / promising / overhyped / vaporware
• Confidence-scored — Single source capped at 6/10
• Action-ready — Commands, configs, rollback steps when applicable
• Security-screened — Prompt injection detection on all content

165+ sources: GitHub, ArXiv, Reddit, HN, X/Twitter, YouTube (with transcripts), RSS, changelogs.

Security

• All actions include requires_user_approval: true
• No executable code is ever sent without explicit approval gates
• Content screened for prompt injection
• Your agent should always present actions to the user before executing

Links

• Website: https://agentwyre.ai
• API docs: https://agentwyre.ai/api/status
• Support: [email protected]