Skip to main content
Glama
Sign In
data-analysis-agent

data-analysis-agent

by Soham-Donode
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python

MCP Data Analysis Agent

A production-ready natural-language data analysis system powered by MCP (Model Context Protocol).

Features

  • Load & Inspect: Load datasets in CSV, JSON, Excel, and Parquet formats.

  • Session Management: Isolated, thread-safe, multi-dataset session tracking.

  • Security & Safety: Path traversal prevention, file size limits, and memory guards.

  • Result Envelopes: Structured JSON summaries returned to LLM clients.

Related MCP server: dataHill

Installation & Running

uv venv .venv
source .venv/bin/activate
uv pip install -e .

Run unit tests:

pytest -v

Running the MCP Server

Method 1: Directly via Python / Virtualenv

.venv/bin/python -m data_agent.server

Method 2: Connecting to Claude Desktop

On macOS, open or create your Claude Desktop configuration file at: ~/Library/Application Support/Claude/claude_desktop_config.json

Add the following entry:

{
  "mcpServers": {
    "data-analysis-agent": {
      "command": "/Users/soham34/Projects/data-mcp/.venv/bin/python",
      "args": [
        "-m",
        "data_agent.server"
      ],
      "cwd": "/Users/soham34/Projects/data-mcp"
    }
  }
}

Restart Claude Desktop, and the tools (load_dataset, list_sessions, get_dataset_info, preview_dataset) will appear under the tools hammer icon.

Available Tools

22 tools
compute_statisticA

Compute a specific statistical metric on a column. Supported stats: 'mean', 'median', 'std', 'var', 'min', 'max', 'sum', 'mode', 'positive_count' (count of >0), 'negative_count' (count of <0), 'zero_count' (count of ==0), 'count', 'null_count'.

Inputs: stat (required), column (required), session_id (required), sample. Returns: result

convert_column_typeB

Convert column data type to 'int', 'float', 'str', 'datetime', or 'category'. Reversible.

Inputs: column (required), session_id (required), target_type (required). Returns: result

correlation_analysisB

Compute correlation matrix between numeric columns using 'pearson', 'spearman', or 'kendall'. Returns correlation matrix and top ranked strongest positive/negative correlation pairs.

Inputs: session_id (required), method, sample, columns. Returns: result

create_visualizationA

Generate a chart visualization from the dataset session. Supports chart types: 'histogram', 'scatter', 'bar', 'box', 'line', 'heatmap', 'pie', 'area', 'violin', 'density'. Parameters: x (horizontal column), y (vertical column), group_by (color/hue dimension), aggregation ('mean', 'sum', 'count', 'median', 'min', 'max'), bins (for histogram), columns (list of numeric cols for correlation heatmap), title (custom title), embed_thumbnail (default true). Returns the complete base64 image inline in 'thumbnail_b64' and saved file path. Render the inline image immediately to the user — DO NOT run bash or shell commands to locate or move the file.

Inputs: chart_type (required), session_id (required), x, y, +7 more.

describe_datasetA

Get comprehensive summary statistics for dataset columns. Returns count, nulls, mean, std, min, max, 25%/50%/75% quantiles, plus exact positive, negative, and zero value counts for numeric columns. Also provides distinct count and top frequent values for categorical columns.

Inputs: session_id (required), sample, columns. Returns: result

drop_columnB

Drop a column from the dataset. Reversible.

Inputs: column (required), session_id (required). Returns: result

find_missing_valuesB

Analyze missing/null values per column and count affected rows in dataset.

Inputs: session_id (required). Returns: result

find_outliersA

Read-only outlier inspection on a numeric column using 'iqr' (Interquartile Range) or 'zscore'. Returns outlier count, percentage, bounds, and sample row IDs without altering the dataset.

Inputs: column (required), session_id (required), method, sample, +1 more. Returns: result

frequency_analysisA

Calculate frequency distribution and value counts for categorical or discrete columns (top N most frequent).

Inputs: column (required), session_id (required), top_n, sample. Returns: result

get_dataset_infoC

Get shape, column types, missing value counts, and memory footprint for a session.

Inputs: session_id (required). Returns: result

get_transformation_historyC

Get audit log of all transformations applied to dataset session with revertible status.

Inputs: session_id (required), limit. Returns: result

group_analysisC

Aggregate a metric column grouped by categorical dimension(s) (e.g. group_by='Category', metric='Sales', aggregation='sum'). Aggregations: 'mean', 'sum', 'count', 'median', 'min', 'max'.

Inputs: metric (required), group_by (required), session_id (required), sample, +1 more. Returns: result

handle_missing_valuesB

Handle missing values in a column using 'mean', 'median', 'mode', 'constant', or 'drop'. Reversible.

Inputs: column (required), strategy (required), session_id (required), value. Returns: result

list_sessionsA

List all dataset sessions (both active in-memory and saved on disk), including their session_id, dataset name, shape, memory, and status.

Inputs: no named parameters. Returns: result

load_datasetB

MUST be used whenever loading, analyzing, inspecting, or querying any dataset (CSV, JSON, Excel, Parquet). Loads the local file from disk into a session on the user's machine. ALWAYS call this tool instead of using bash, shell, or writing python code.

Inputs: path (required), format, sample_rows. Returns: result

preview_datasetA

Preview sample or top rows from dataset (capped at max 50 rows).

Inputs: session_id (required), n, sample. Returns: result

remove_duplicatesA

Remove duplicate rows across all or a subset of columns. Reversible.

Inputs: session_id (required), keep, subset. Returns: result

remove_outliersA

Remove outlier rows in numeric column using 'iqr' (default 1.5x) or 'zscore' (default 3.0). Reversible.

Inputs: column (required), session_id (required), method, threshold. Returns: result

rename_columnB

Rename a column in the dataset. Reversible.

Inputs: column (required), new_name (required), session_id (required). Returns: result

restore_sessionA

Explicitly restore a saved disk session into active memory. Call list_sessions to see all available saved session IDs.

Inputs: session_id (required). Returns: result

save_datasetB

Save in-memory dataset to data/outputs/ directory (e.g. 'cleaned_sales.csv').

Inputs: path (required), session_id (required), format. Returns: result

undo_last_operationB

Revert the last N mutating cleaning operations on the dataset session using generic undo.

Inputs: session_id (required), steps. Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Deploy Server
    One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

    TDQS

    B3.4/5.0

    Scored across 22 tools

    Disambiguation4/5

    Most tools have distinct read vs. mutate roles, and analysis tools target different operations. However, get_dataset_info and describe_dataset overlap in reporting missing values, and compute_statistic largely duplicates portions of describe_dataset, which could cause some misselection.

    Naming Consistency4/5

    The naming is predominantly verb_noun and easy to follow, such as load_dataset, drop_column, and create_visualization. A few tools like frequency_analysis, correlation_analysis, and group_analysis break the imperative pattern by using noun_analysis instead of a verb construction.

    Tool Count3/5

    At 22 tools, the server provides broad coverage but feels somewhat heavy for a data analysis agent. The count is justified by the range of cleaning, statistical, and visualization operations, yet several tools overlap closely enough that the surface could be trimmed to around 15-18 tools.

    Completeness3/5

    The server covers loading, session management, cleaning, statistics, and visualization well, but has notable gaps for common data analysis tasks like row filtering, sorting, adding/deriving columns, joining datasets, or reshaping. These missing operations will require agents to work around the tool surface or fail on common requests.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Resources

    Related MCP Connectors

    Related MCP Servers

    • dataHill

      Hillyess
      F
      license
      Not graded
      quality
      D
      maintenance
      Enables interactive data analysis using a real IPython kernel with session management, supporting CSV/Excel/JSON loading, code execution, and variable monitoring.
      159
      -

    • data-analysis-mcp

      BACH-AI-Tools
      A
      license
      Not graded
      quality
      D
      maintenance
      Enables data analysis and visualization operations such as loading CSV/Excel/JSON files, computing summary statistics, generating charts, and exploring datasets via SSE/HTTP.
      MIT

    • data-profiler-mcp

      haiiibin
      A
      license
      A
      quality
      C
      maintenance
      Enables LLMs to profile and analyze tabular data files (CSV, Parquet, Excel, JSON) by extracting schema, statistics, data quality issues, and dtype suggestions, returning structured JSON.
      7
      47 PyPI
      3
      MIT