Cliffwatch

Local, read-only Solana memecoin intelligence terminal that scores every live token in real time.

Cliffwatch watches live token feeds and scores every token with Risk/Scam, Hype, and conservative Gem points using many public data sources, local algorithms, and optional local LLM summaries. It enriches tokens with scam/rug/wallet/market signals and summarizes evidence without connecting a wallet or trading.

Download

Package	Link
Windows EXE archive	Download
Local build output	dist\Cliffwatch-latest.zip after .\scripts\build_exe.ps1

What It Does

Area	Features
Token feed	Live DEX Screener discovery, optional Pump.fun/PumpAPI sources, search, favorites
Risk checks	Scam Intel Triage, RugCheck, Solana authority checks, holder concentration
Market intel	Liquidity, volume, transaction momentum, dynamic trend arrows
Scoring	Real-time Risk/Scam, Hype, and risk-gated Gem points from local scoring algorithms
Wallet analysis	Wallet-cluster view and on-demand parsed wallet-flow scan
Evidence	Compact evidence watch cards, scam/rug/community/market summaries
Local-first	DuckDB storage, no wallet connection, no signing, no buying or selling

Scoring Model

Cliffwatch uses three separate 0-100 scores:

Score	Meaning
Rug	Red-flag and rug-risk score. Higher means more danger from scam intel, RugCheck, token authorities, holder concentration, wallet clusters, or suspicious token controls.
Hype	Market momentum and attention score. Higher means stronger volume, liquidity, transaction activity, buy pressure, profile presence, and social surface. High Hype is not proof of safety.
Gem	Risk-gated buy-quality score. It combines Hype with low red flags, safer token controls, holder distribution, social quality, and scam-intel clearance. High Gem means the coin is a stronger candidate for manual review.

The feed highlights high-Gem coins with a green outline. Clean Gem rows require high Gem, low Rug, complete analysis, and clean or verified scam-intel status. Hover over Rug, Hype, or Gem badges in the app for a short metric reminder.

Quick Start

For most users, run the Windows EXE.

1. Get Cliffwatch-latest.zip.
2. Unzip it.
3. Double-click Cliffwatch.exe.

Cliffwatch starts a local app and opens your browser automatically. If the browser does not open, use the URL printed in the EXE window, usually:

http://localhost:9000/

Close the EXE window or press Ctrl+C to stop Cliffwatch.

If you cloned this repo and a local build already exists, you can run:

.\dist\Cliffwatch.exe

For local source runs, use the app command shortcuts:

.\start-app.cmd

Rebuild The EXE

Use these steps only if you are changing the app or need to create a fresh EXE from source.

Copy-Item .env.example .env -ErrorAction SilentlyContinue
.\scripts\build_exe.ps1

Outputs:

dist\Cliffwatch.exe
dist\Cliffwatch-latest.zip

Run the rebuilt app with:

.\dist\Cliffwatch.exe

The build script creates the Python virtual environment if needed, installs the API dependencies and PyInstaller, packages the local static app, and writes the release archive. It also creates build\Cliffwatch-release as a temporary staging folder for the zip; after the archive is built, use dist\Cliffwatch.exe or dist\Cliffwatch-latest.zip as the canonical outputs.

Run From Source

Use source mode for development only. The easiest command is:

.\start-app.cmd

For first-time setup or manual development startup:

Copy-Item .env.example .env -ErrorAction SilentlyContinue
py -m venv .venv
.\.venv\Scripts\python -m pip install -r apps\api\requirements.txt
.\scripts\dev.ps1 -AutoPort

Open the URL printed in the terminal.

Stop And Restart

Use these commands from the repo root:

.\start-app.cmd
.\stop-app.cmd
.\reload-app.cmd

start-app.cmd starts the API and web app with automatic port selection. stop-app.cmd stops the tracked Cliffwatch processes. reload-app.cmd stops and starts the app again.

Optional Scam Intel APIs

Static scam-intel lists work without API keys. For stronger live triage, set any of these optional keys in .env:

CLIFFWATCH_JUPITER_API_KEY=
CLIFFWATCH_CHAINABUSE_API_KEY=
CLIFFWATCH_GOPLUS_ACCESS_TOKEN=
CLIFFWATCH_CHAINPATROL_API_KEY=
CLIFFWATCH_BLOCKSEC_API_KEY=
CLIFFWATCH_HELIUS_API_KEY=
CLIFFWATCH_INTELX_API_KEY=

Helius is used as the primary Solana RPC when configured, with public Solana RPCs kept as fallback. BlockSec MetaSleuth adds address risk-score rows to scam-intel triage. Intelligence X adds low-confidence selector context for token domains and URLs. Etherscan and Blockchair keys are accepted in settings for future cross-chain flows, but they are not used in the Solana scoring path yet.

Each live API has short timeouts, rate limits, and caching so the stream can keep processing tokens when an optional provider is slow or unavailable.

Local LLM

Cliffwatch can use Ollama for local summaries.

1. Install Ollama: https://ollama.com/download
2. Pull the model:

ollama pull qwen3-vl:8b-orion-instruct

Set .env:

CLIFFWATCH_OLLAMA_MODEL=qwen3-vl:8b-orion-instruct

Tests

python -m pytest apps\api\tests