WaveGuard MCP server

Find outliers and anomalies in structured data — ideal as a second step after pulling records from Google Sheets, Airtable, Supabase, Notion databases, HubSpot, Financial APIs, GitHub, NPM, or any source that returns rows of JSON. Fully stateless: send known-good rows as training and suspect rows as test in ONE call. Returns per-row anomaly scores, confidence levels, and the top features explaining WHY each row was flagged. Typical workflow: (1) Pull data from another tool (e.g. Google Sheets, Supabase query, HubSpot deals). (2) Pass the first N rows as training (normal baseline). (3) Pass remaining or new rows as test. (4) Report which rows are anomalous and why. Works on JSON objects, numbers, text, arrays. No separate training step required. Examples: - Spreadsheet QA: Pull 500 sales rows from Sheets → train on first 400 → test last 100 → flag outlier entries - Financial screening: Get ratios for 50 stocks from a financial API → find anomalous ones - CRM hygiene: Pull HubSpot deals → flag deals with unusual discount/value patterns - Dependency audit: Get NPM package metrics → flag packages with anomalous quality scores - Commit review: Pull GitHub commit metadata → flag unusual commit patterns

Detect anomalies in time-series data — use after pulling numeric metrics from monitoring APIs, financial data sources, IoT sensors, or spreadsheet columns. Send a single numeric array and specify a window size. Early windows define 'normal', recent windows are tested for anomalies. Typical workflow: (1) Pull a column of numbers from Sheets, a Supabase time-series table, or a metrics API. (2) Pass the array here. (3) Get back which time windows are anomalous. Examples: - Revenue monitoring: Pull monthly revenue from Sheets → detect anomalous months - Stock screening: Pull 90 days of closing prices → find unusual price windows - Server health: Pull response-time metrics → identify degradation windows - Sensor QA: Pull temperature readings from IoT API → flag sensor drift

Check WaveGuard API health, GPU availability, version, and engine status. No authentication required. Returns status, version, and GPU info.

Get a physics embedding of any data item (52-dim at Level 0, 62-dim at Level 1 with phase statistics). The fingerprint captures structural properties via wave-equation dynamics — useful for similarity search, clustering, baseline comparison, and drift detection. Works on JSON objects, token metrics, wallet activity, trading data, or any structured data. Returns a deterministic vector with labeled dimensions (chi statistics, energy distribution, gradient patterns, and phase coherence at Level 1).

Compare two data items for structural similarity using physics-based fingerprints. Returns cosine similarity (0–1) and Euclidean distance. Use for duplicate detection, behavioral matching, drift analysis, or checking if two tokens/wallets/contracts are structurally similar. Cosine similarity > 0.95 = very similar. < 0.80 = structurally different.

Assess crypto token legitimacy risk. Send metrics from known-good tokens as training (price, volume, holders, liquidity, market_cap, age_days, etc.) and suspect tokens as test. Detects pump-and-dump patterns, fake metrics, and anomalous token profiles. Example: Pull CoinGecko data for 20 established tokens → train. Test a new token → get risk score and which metrics are suspicious.

Profile wallet behavior against baselines. Send normal wallet transaction patterns as training (tx_count, avg_value, unique_tokens, gas_spent, active_days, etc.) and suspect wallets as test. Detects bot activity, wash trading wallets, and sybil patterns. Example: Profile 50 organic wallets → test 10 suspect addresses.

Detect wash trading and fake volume in OHLCV candle data. Send known-legitimate candles as training and suspect candles as test. Detects artificial volume spikes, suspiciously regular patterns, and manipulated price-volume relationships. Example: Send 100 candles from a liquid pair as baseline, test candles from a suspicious pair.

Detect price manipulation in time-series data. Send a price or price+volume history as a numeric array. Early windows define 'normal' trading, recent windows are tested for manipulation patterns (pump-and-dump, spoofing, layering). Example: Send 90 days of closing prices → detect manipulated windows.

Fetch live crypto market data from CoinGecko and DexScreener. No external data needed — WaveGuard pulls it for you. Use 'coin_id' for CoinGecko (e.g. 'bitcoin', 'ethereum', 'solana'). Use 'contract_address' for DexScreener (any chain). Use 'search' to find token IDs by name/symbol. Returns: price, volume, market cap, liquidity, price history, OHLC candles — ready to feed into waveguard_token_risk, waveguard_volume_check, or waveguard_price_manipulation.

Run baseline plus counterfactual variants and measure verdict/score sensitivity.

Analyze sequence drift and regime shifts over ordered samples.

Estimate instability under controlled perturb-and-resolve trials.

Measure coherence/entropy and collapse-risk indicators for candidate data.

Compute pairwise interaction matrix and cluster decomposition for entities.

Estimate shock propagation and resilience from adjacency-linked entities.

Run targeted interventions and rank effect sizes.

Score candidate actions and extract robust action zones.

Compute horizon-specific anomaly outlook and consistency across windows.