xrpldashboard

Public XRPL analytics with a built-in truth-audit system. Live at xrpldashboard.com.

What this is

Three things make this repo different from "another blockchain dashboard":

1. Every headline number has a claim record. CLAIMS.yaml enumerates the public numeric claims across /rlusd, /whales, /coverage, and /analytics, each tied to its source function and freshness contract. scripts/claims_check.sh runs before pushes.
2. A four-layer audit catches stale numbers before readers do. See docs/TRUTH_AUDIT_DESIGN.md. Retro-tested against the last nine real incidents: 9 of 9 caught. The first live catch (an RLUSD partial-day bug) surfaced within hours instead of the 53 days it had previously gone unnoticed.
3. Daily signed snapshots. Ed25519-signed, Merkle-chained. /signed-snapshots and .well-known/snapshots/<date>.json let anyone verify a number as-of a date without trusting the site.

Quick links: /methodology · /signed-snapshots · docs/TRUTH_AUDIT_DESIGN.md

Agent tier (MCP)

Live public MCP endpoint: https://mcp.xrpldashboard.com/mcp — streamable-HTTP, protocol version 2025-06-18, no auth. Fifteen read-only tools over XRPL and the on-XRPL RLUSD supply, every response wrapped in a proof-annotation envelope with source, as_of, freshness_contract, and a claims_ref back to /claims. Public beta through 2026-09.

Connect in 60 seconds — copy-paste config for Claude Desktop or any MCP client:

{
  "mcpServers": {
    "xrpldashboard": {
      "command": "npx",
      "args": [
        "mcp-remote@latest",
        "https://mcp.xrpldashboard.com/mcp"
      ]
    }
  }
}

Or add it as a Custom Connector in Claude Desktop → Settings → Connectors with URL https://mcp.xrpldashboard.com/mcp and auth None. Full onboarding page (three sample prompts, honest limits, 429 shape): /connect.

• Session rate limit: 600 tool calls / hour / session, enforced live (HTTP 429 + Retry-After on breach). See mcp_session_rate_limit.py.
• Discovery manifest: /.well-known/agents.json · /llms.txt · /openapi.json.
• Verifiable moat: get_signed_snapshot + verify_snapshot_signature — pin the Ed25519 pubkey at /.well-known/snapshots/pubkey.pem and verify a day's snapshot without trusting us.
• Backing infra: our own rippled full-history node (Ubuntu box in Indiana → Cloudflare Tunnel), source at mcp_server.py and mcp_tools_*.py.

Design doc: docs/AGENT_TIER_DESIGN.md.

Pages

37 public pages, organized as:

• Money flow: /whales, /pools, /tokens, /token/<id>, /mpts, /mpt/<id>, /rlusd, /lending
• Institutional & regulatory: /institutional, /regulation, /rwa, /credentials, /amendments, /sidechain
• Coverage & audit: /coverage, /methodology, /signed-snapshots, /verify, /walker-health, /health
• Reader tools: /check, /wallet/<id>, /network, /learn, /price-data, /help/already-sent-money
• Trust & meta: /about, /security, /privacy, /terms, /subprocessors, /contact

Architecture

• Web: Flask 3.1 + Jinja2 + Flask-Babel (i18n) + Flask-Limiter, deployed on Render, fronted by Cloudflare.
• Data: Neon Postgres (single shared DB across web + walkers).
• Ingest: 29 background services under launchd — 23 ingest walkers and canaries, plus 4 backup and 2 snapshot-signing jobs. Each walker writes to walker_health; /walker-health surfaces stalls.
• XRPL client: xrpl-py 4.5.0 with a local-first cascade to public rippled nodes; silent-failover attempts logged to walker_node_fallback so we can audit reliability rather than hope.
• Signing: Ed25519 via cryptography (see signed_snapshot.py).

The truth-audit system (why this repo may be reusable)

Most public dashboards trust their own writes. This one doesn't. Four layers, working together:

• Layer 1 — walker health. walker_health rows + /walker-health page. Catches "the number stopped moving because the writer died."
• Layer 2 — plausibility rules. Continuous checks like "24h net-change should not equal zero for 53 days." Catches "the number is moving in the DB but the query is wrong."
• Layer 3 — external cross-check. Third-party comparisons (e.g., independent Ethereum RPCs for RLUSD supply, CoinGecko for XRP price) surface disagreement on the same measurement.
• Layer 4 — claims manifest. CLAIMS.yaml + scripts/claims_check.sh. Every headline number is enumerated with its source function and freshness contract; the checker exits non-zero on drift.

Full design and the 9-of-9 retro-test are in docs/TRUTH_AUDIT_DESIGN.md.

If you're building a public-data project and want the pattern, the design doc is written to be lifted.

Running locally

What runs in five minutes: the Flask web app against a copy of the schema. Enough to click through pages and see the UI.

What does not run in five minutes: the full site, because a live dashboard needs the walker fleet ingesting from XRPL against Neon Postgres. Standing up the walkers is a real operations task, not a docker compose up.

Quickstart (web app only):

python3 -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env      # then fill in DATABASE_URL, FLASK_SECRET_KEY, etc.
python app.py             # http://localhost:5001

For the full ingest setup, see docs/TRUTH_AUDIT_DESIGN.md and the launchd/ plists (which are the author's local install — forkers will want to path-adjust).

Independence

xrpldashboard is not operated by Ripple, the XRPL Foundation, or any exchange. No paid placements. No affiliate links on labeling or metric surfaces.

Funding

Development is self-funded, with support from community grant programs where they align with the mission (public-goods data infrastructure for XRPL).

Contributing

Issues and PRs welcome. If you're touching a page that renders a headline number, add or update the corresponding entry in CLAIMS.yaml; scripts/claims_check.sh will remind you.

License

MIT. See LICENSE.

Disclaimer

xrpldashboard reports on-chain data and public regulatory information. Nothing on the site or in this repo is financial, legal, or investment advice.