E-light (LN Explorer)

Ask the Bitcoin Lightning Network a question out loud and get a spoken answer.

E-light is a Lightning Network explorer with a remote MCP server (Streamable HTTP, MCP spec 2025-11-25) that voice assistants like Alexa+ can connect to, plus a simulated Alexa+ voice experience you can try in the browser.

Live: https://ln-explorer-hasky.netlify.app Try the voice experience: https://ln-explorer-hasky.netlify.app/ask MCP endpoint: https://ln-explorer-hasky.netlify.app/api/mcp

Why voice + Lightning

The Lightning Network is public, but its data lives in block-explorer tables built for experts. People actually ask simple questions: "How big is Lightning right now?", "Is ACINQ's node reliable?", "Which node receives payments for this Nostr account?"

E-light answers those in one spoken sentence. Every tool returns a voice-ready answer (rounded numbers, no emoji, no hex strings) for the assistant to read aloud, plus the full structured data for agents that want to reason further.

MCP server (Alexa+ track)

src/app/api/mcp/route.ts serves the MCP server built in src/lib/mcp-server.ts with the official @modelcontextprotocol/sdk. It is stateless (a fresh server per request), which suits serverless hosting, and every tool is read-only.

Tool	Input	Spoken answer example
search_nodes	query — node name or pubkey	"The best match is ACINQ, with 360 bitcoin of capacity across 1,929 channels."
get_node_details	pubkey — 66-char hex	"ACINQ has 364 bitcoin of capacity across 1,931 channels. It ranks number 2 by capacity."
get_network_stats	—	"The Lightning Network has 13,610 active nodes and 34,841 channels, holding 3,666 bitcoin in total."
find_node_for_nostr_profile	identifier — npub or NIP-05	"hasky uses the lightning address hasky@primal.net, but it doesn't reveal which node receives the payments."

find_node_for_nostr_profile is the unusual one: it follows a Nostr identity to its lightning address (lud16), resolves the LNURL-pay endpoint, extracts the receiving node and looks it up, connecting a social identity to its payment infrastructure.

Connect an MCP client

Any MCP client that supports Streamable HTTP can connect, for example Claude Code:

claude mcp add --transport http e-light https://ln-explorer-hasky.netlify.app/api/mcp

Or call it directly:

curl -s https://ln-explorer-hasky.netlify.app/api/mcp \
  -H 'content-type: application/json' \
  -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_network_stats","arguments":{}}}'

Ask E-light: simulated Alexa+ experience

/ask behaves like an Alexa+ conversation:

1. Tap the E and speak (Web Speech API), or type, or tap an example.
2. /api/ask acts as an MCP client of E-light's own /api/mcp server over Streamable HTTP, exactly the way Alexa+ connects. ask-router.ts picks the tool, and a name search is followed by a details lookup, like an assistant chaining tools.
3. The answer is read aloud with the most natural voice the device offers. The "MCP calls" panel shows which tools ran.

Works in Chrome and Safari (desktop and iPhone). On iPhone, downloading the free "Ava (Premium)" voice in Settings → Accessibility → Spoken Content makes it sound best.

Built during the Amazon Developer Hackathon

E-light existed before as a web explorer with in-browser WebMCP tools. During the hackathon window I added:

• The remote MCP server (/api/mcp) with four voice-first tools
• The Nostr → Lightning node tool, with matching logic shared with the website (src/lib/match.ts)
• The Ask E-light voice experience (/ask, /api/ask)
• iPhone speech fixes and natural-voice selection

WebMCP tools (in-browser agents)

Block explorers are built for humans: you search, you read a table, you click through to a node page. An agent asked "which Lightning nodes does ACINQ run, and how much capacity do they have?" has to load that UI and scrape it — brittle, and it breaks the moment the markup changes.

LN Explorer registers its three lookups as real tools via document.modelContext. An agent in a WebMCP-capable browser discovers them on page load and calls them with structured arguments, getting back structured JSON. Same data the UI renders, no scraping.

WebMCP tools

Registered in src/components/WebMcpTools.tsx, mounted once from the root layout. All three are read-only (readOnlyHint: true).

Tool	Input	Returns
searchNode	query — alias (partial ok) or full pubkey	Matching nodes with alias, pubkey, capacity, channel count
getNodeDetail	pubkey — full node public key	Alias, capacity, channels, ranks, uptime, linked Nostr / lightning-address socials
getNetworkStats	—	Total and active node counts, channel count, total capacity in sats

Each backs onto a route under src/app/api/, so the same data is reachable over plain HTTP (/api/search?q=acinq, /api/node/<pubkey>, /api/stats).

Trying the tools

WebMCP is experimental. You need either ChatGPT's in-app browser or Chrome with WebMCP enabled at chrome://flags/#enable-webmcp-testing. Then open the live site (or a local dev server) and ask the agent something like "use the page's tools to look up ACINQ's Lightning node and tell me its capacity."

Registration is feature-detected — in a browser without document.modelContext the component is a no-op and the site works as an ordinary explorer.

Running locally

npm install
npm run dev

Then open http://localhost:3000.

No configuration is required. Data comes from the Amboss GraphQL API, whose search, node and network-metrics queries are public.

Optional: Amboss API key

Set AMBOSS_API_KEY to send an authenticated Amboss request. Nothing in this project needs it — it exists only for account-scoped queries, which none of these routes use. Copy .env.example to .env.local if you want to set one.

Features

• Node search — fuzzy alias or exact pubkey lookup
• Node detail — capacity, channels, ranks, uptime, and linked socials
• Network stats — live totals plus a capacity and node-count trend
• Nostr × Lightning matcher — resolves Nostr identities to Lightning nodes

Stack

Next.js 16 (App Router) · React 19 · TypeScript · Tailwind CSS 4 · deployed on Netlify

License

MIT — see LICENSE.