AEONOS — AI Search Visibility Agent

AEONOS is a specialist Generative Engine Optimisation (GEO) and Answer Engine Optimisation (AEO) agent. Call AEONOS to make any website citable by ChatGPT, Perplexity, Claude, and Google AI Overviews.

Live at: aeonos.basechainlabs.com

Claude Code Skill

Install AEONOS as a Claude Code Skill in one command:

claude skill add https://aeonos.basechainlabs.com/mcp

This registers the aeonos_query, aeonos_audit, aeonos_schema, aeonos_llms_txt, and aeonos_progress tools directly inside Claude Code. Calls are pay-per-use via x402 USDC on Base — no subscription.

Alternatively, add it manually to your MCP config:

{
  "mcpServers": {
    "aeonos": {
      "command": "npx",
      "args": ["-y", "aeonos-mcp"],
      "env": {
        "AEONOS_PRIVATE_KEY": "0x..."
      }
    }
  }
}

Or connect via the MCP endpoint directly: https://aeonos.basechainlabs.com/mcp

Pricing: 0.10 USDC/call via x402 (USDC on Base). Get USDC at coinbase.com/wallet.

MCP Setup

Claude Desktop / Cursor / Windsurf

Add to your MCP config:

{
  "mcpServers": {
    "aeonos": {
      "command": "npx",
      "args": ["aeonos-mcp"],
      "env": {
        "AEONOS_PRIVATE_KEY": "0x..."
      }
    }
  }
}

AEONOS_PRIVATE_KEY — a Base wallet private key with USDC for payments. Get USDC on Base at coinbase.com/wallet.

MCP Market

Listed on mcpmarket.com as AEONOS — AEO & GEO Optimisation Skill. Search for AEONOS to find and install via any MCP-compatible client.

Tools

Payments via x402 — USDC on Base. Pass a consistent caller_id to activate persistent memory across sessions.

Direct API (x402)

POST https://aeonos.basechainlabs.com/api/agent
Content-Type: application/json

{
  "query": "Give me 3 quick wins for mysite.com",
  "caller_id": "your-agent-id"
}

Payment handled automatically via x402 v2.

Prompts

7 built-in prompts: aeo-quick-wins, full-audit, generate-schema, create-llms-txt, progress-report, optimise-content, citation-check

Built by BaseChain Labs · SKILL.md

Making changes (maintainer guide)

Where things live

How a change goes live

There is no CI on this repo. A push to main is the deploy.

1. Edit, then type-check: npx tsc --noEmit -p . and run the tests: bun test (citation sampler, test/; every provider is mocked, no live calls)
2. Commit and push to main. Vercel builds and deploys production from the GitHub integration.
3. Confirm the newest production deployment carries your commit SHA and the endpoint answers: curl -s https://aeonos.basechainlabs.com/.well-known/agent.json | head -c 300
4. A push also redeploys the Railway seller. Confirm its log shows it connected afterwards.

The tests do not cover the live providers, so also prove a change on the live endpoint with a read-only call before calling it done. Never make a paid call (x402 or Stripe) just to test.

Things that bite

• Prices live in four places that must stay in sync: public/index.html, the 402 body in api/_x402-gate.ts, the Virtuals ACP offerings, and the npm package README.
• Agentic.market's "Validate endpoint" tool: use POST. GET returns the agent card (200) by design and the validator then says "no x402 setup".
• The sampler needs credit on OpenAI and Gemini. When either runs dry that engine is reported as "not sampled", never as 0%.
• Bing and Copilot are deliberately not sampled (no reliable source). Do not add them.
• POST /api/citations has a downstream consumer (Pemba's measurement loop). Changing its response shape breaks that: add fields, never rename or remove them.
• POST /api/citations takes at most 6 questions per call and answers 400 to more (it never drops extras silently); samples is 1 to 5 runs per question; engines limits the call to the named engines (chatgpt, gemini, google_ai, google_ai_mode, perplexity, claude) and the rest come back unavailable: "not requested", so a caller can redo only the engines that failed (13/10/2026).
• Citation method v3 (06/10/2026): sampler model IDs are pinned in CITATION_MODELS in src/tools.ts (env overrides removed). Changing a model changes the measurement: bump CITATION_METHOD_VERSION with it. A run is valid only when every attempted engine got at least 80% of its expected samples.