MCP Analysis

Analysis engine for the MCP ecosystem. Pipeline-based data analysis with pluggable functions, transforms, and multi-source data integration. Implements the AnalysisPort / AnalysisFunctionPort / AnalysisDataSourcePort Contract Layer defined in mcp_bundle.

Components

  • Spec — spec validator, parameter resolver, spec manager.
  • Artifact — artifact builder, store, provenance tracker.
  • Execution — execution engine, batch / ad-hoc / stream executors, job manager, retry policy, step logger.
  • DataSource — multi-source data integration with pluggable adapters, including a synthetic source for simulation: signal models, linear dynamic systems (state-space / transfer function / RLC, RK4), and Monte Carlo ensemble prediction bands.
  • Functions and Transforms — 36 built-in functions: statistics (incl. moments, PCA, hypothesis tests), DSP (FFT w/ phase, Welch PSD, spectrogram, cepstrum, harmonics/THD, cross-PSD/coherence/FRF, filters incl. notch/Butterworth-cascade/median, software lock-in, peaks, resample, envelope), timeseries (ACF auto-period, cross-correlation, CUSUM, Holt-Winters, Savitzky-Golay w/ derivatives, Lomb-Scargle for uneven sampling, Kalman filter/smoother), interpolation, and a built-in domain layer (vibration ISO 10816, HRV, EEG bands) — plus pluggable custom functions and transforms.
  • Alerts — alert evaluation against analysis results.
  • MCP integration — exposes analysis capabilities through MCP tools.
  • Standard port adapter — analysis_port_adapter.dart implementing the mcp_bundle analysis Contract Layer.

Quick Start

import 'package:mcp_analysis/mcp_analysis.dart';

final engine = ExecutionEngine(
  jobManager: JobManager(),
  artifactStore: InMemoryArtifactStore(),
);

final result = await engine.executeAdhoc(spec, parameters);

Support

License

MIT — see LICENSE.

Libraries

mcp_analysis
MCP Analysis package.