Create validated, maintainable data charts and dashboards with Vizro tools and templates.
Data Science Tools
Data science servers connect analysis environments, computation and datasets to MCP clients. Check supported formats, where code executes and whether the integration matches your existing notebook or analysis workflow.

Convert almost any file or web content into Markdown.
Interact with Jupyter.
Data quality evaluation with Dingo: run rule-based and LLM-based evaluations and list its rules and prompts.
Autonomous data exploration of CSV datasets with generated insights.
Interact with and control Jupyter Notebooks.
Forecasting and prediction with Chronulus AI agents.
Orchestrate hyperparameter search and other optimization tasks with Optuna.
Link multiple data sources (SQL, CSV, Parquet and more) and analyze the data for insights and visualizations.
Connect to Kaggle to download and analyze datasets.
Create, manage and automate Label Studio projects, tasks and predictions for data labeling workflows.
Connect AI agents to interactive live R sessions.
Create and interact with GrowthBook feature flags and experiments.
Graph analysis with NetworkX: centrality algorithms, community detection, PageRank and graph visualization.
Math engine combining SymPy, NumPy and Matplotlib for symbolic algebra, numerical computing and data visualization.
Query local information about dependencies in a Ruby project's Gemfile.
Math tools for agents: theorem search, PSLQ constant identification, OEIS lookup, Lean kernel checks and applicability checklists.
Decision intelligence algorithms: bandits, Monte Carlo, constraint optimization, forecasting, anomaly detection, risk analysis and graph algorithms.
Code refactoring engine: code analysis, hotspot detection, complexity metrics, persistent memory and automated refactoring plans.
Mathematical computations: basic arithmetic, advanced functions, statistical analysis, unit conversions and financial calculations.
Judgment calls on small, cheap models: yes/no, pick-one and rate questions about any text, answered YES, NO or UNSURE. Runs on TypeSafe's Jev, open models on Ollaya, Ollama or a local NLI model; each judges up to 1,000 texts per call.
Exploratory data analysis that finds feature interactions and subgroup effects, with p-values, effect sizes and literature citations.
Transparent statistical analysis with cited findings, integrity-hashed result bundles and a calibration corpus of measured detector false-fire rates.
Educational server for math operations, statistics, visualization and persistent workspaces.
LLM quantization: convert models to GGUF, GPTQ and AWQ, recommend quantization settings, evaluate quality and push to Hugging Face Hub.
Offline tabular data access: describe, preview and query CSV, TSV, JSON, Parquet and Excel files, diff datasets and build an interactive HTML explorer.
Create, read, validate and save Stella system dynamics models (.stmx files in XMILE format) for scientific simulation and modeling.
Physics-based anomaly detection with Klein-Gordon wave equations: scans, fingerprinting, token risk, wallet profiling and price manipulation detection.
Browse Kaggle competitions, leaderboards, models, datasets and kernels.
Profile local tabular files (CSV, TSV, Parquet, Excel, JSON): dataset overview, per-column statistics, data-quality audits and dtype suggestions.
Offline exact symbolic math via the Giac/Xcas computer algebra system: calculus, equation solving, linear algebra, combinatorics and claim verification.
Real-time LLM and VLM comparison with benchmarks, prices and personalized recommendations from multiple data sources.
Data compression with gzip, Brotli, deflate and TurboQuant quantization, with automatic algorithm selection.
Structural observability for AI conversations: detects loops, stuck states, breakthroughs and convergence without analyzing content.
Numerical methods: root finding, numerical integration, differential equations (RK4) and Lagrange interpolation.
Compare approximate filter data structures: Bloom, Counting Bloom, Cuckoo and SuRF.
Deterministic date, math and unit tools: time snapshots, calendar math with natural-language parsing, expression evaluation and Pint-based unit conversion.
Classical hypothesis testing, effect sizes, power and sample-size analysis and multiple-comparisons correction, returned as cited, assumption-checked answers.
ML experiment contracts (cq.yaml plus JSON) to run, validate, compare and trace experiments in any framework, including PyTorch, Lightning and sklearn.
Offline unit conversion and dimensional analysis backed by the GNU units database, with compound expressions and reduction to SI base units.
Fast, local, non-autoregressive System-1 decision engine as MCP tools β typed decisions (classify / score / yes-no) over text, email, tickets or JSON in 100+ languages, with calibrated confidence and no text generation, so nothing to hallucinate. Offloads routing/triage/gating judgment calls from the main model; runs on-device, offline by default. pipx install mcp-laya
Deterministic multi-criteria decision analysis: rank options on weighted criteria via weighted-sum, weighted-product or TOPSIS, with sensitivity analysis.
Local deterministic calculations: exact-precision math, date arithmetic, unit conversion, statistics, subnet math, haversine and Levenshtein distance.
QUBO/Ising optimization with classical, hybrid and quantum solvers, paid per call via x402 (USDC on Base).
Probabilistic time-series forecasting: quantile forecasts from Chronos-2, TimesFM 2.5, Toto 2, TiRex-2 and IBM Granite foundation models, routed to one model, named explicitly, or ensembled. Remote server with an npm stdio bridge.
Run zero-shot Jev AI typed decisions with criteria defined in the request and get calibrated probabilities for every option.