data-cleaner-agent MCP server
✓ VerifiedCleans messy CSV files using an agentic workflow where an LLM plans transformations and tested Python functions execute them without touching raw data values.
Also known as wesseltl/data-cleaner-agent, agentic-csv-cleaner
- Stars
- 0
- Tools
- —
- Downloads, 30 days
- 355
- Latest version
- 0.1.3
- Last commit
- Aug 10, 2026
- License
- MIT
What is data-cleaner-agent MCP?
data-cleaner-agent cleans messy CSV files using an agentic workflow where an LLM plans transformations and tested Python functions execute them without touching raw data values.
Summary generated from the project's README
About data-cleaner-agent MCP server
Cleaning messy CSV files through an agentic workflow is what this server enables for developers and AI agents. It uses a planning system to decide which transformations are necessary, such as fixing headers, trimming whitespace, or standardizing dates and currencies, while using tested Python functions to execute the actual changes. This separation ensures that a language model never touches or rewrites your raw data values directly, preventing hallucinations or silent data corruption.
Users can run the server locally using a RuleBasedPlanner that relies on offline heuristics without any API keys. For more complex tasks, you can install the LLMPlanner by providing an ANTHROPIC_API_KEY to allow a model to generate a JSON plan. The server integrates with MCP clients like Claude Desktop, Cursor, Windsurf, and GitHub Copilot Agent mode. It provides a detailed report for every operation, distinguishing between successful parses and values that could not be converted, such as unparseable numbers or unmapped categories.
What you can do with it
- 1Standardize inconsistent date formatsAsk the agent to convert various date strings into a single ISO format.
- 2Clean currency and numeric columnsHave the tool transform mixed currency symbols and European decimal formats into standard numbers.
- 3Remove duplicate data entriesInstruct the AI to identify and delete redundant rows from a dataset.
- 4Validate data cleaning resultsUse the generated report to identify which specific cells failed to parse correctly.
Written from the project's README. Every use case is checked against a quote from it.
How to install data-cleaner-agent MCP
Formats verified 10 Nov 2026Add to CursorAdd to GooseInstall data-cleaner-agent MCP in Claude Code
$ claude mcp add data-cleaner-agent -- uvx agentic-csv-cleanerInstall data-cleaner-agent MCP in Cursor
~/.cursor/mcp.json
{
"mcpServers": {
"data-cleaner-agent": {
"command": "uvx",
"args": [
"agentic-csv-cleaner"
]
}
}
}Install data-cleaner-agent MCP in VS Code
.vscode/mcp.json
{
"servers": {
"data-cleaner-agent": {
"type": "stdio",
"command": "uvx",
"args": [
"agentic-csv-cleaner"
]
}
}
}Install data-cleaner-agent MCP in VS Code CLI
$ code --add-mcp '{"name":"data-cleaner-agent","type":"stdio","command":"uvx","args":["agentic-csv-cleaner"]}'Install data-cleaner-agent MCP in Claude Desktop
claude_desktop_config.json
{
"mcpServers": {
"data-cleaner-agent": {
"command": "uvx",
"args": [
"agentic-csv-cleaner"
]
}
}
}Install data-cleaner-agent MCP in Codex
$ codex mcp add data-cleaner-agent -- uvx agentic-csv-cleanerInstall data-cleaner-agent MCP in Codex config.toml
~/.codex/config.toml
[mcp_servers.data-cleaner-agent] command = "uvx" args = ["agentic-csv-cleaner"]
Install data-cleaner-agent MCP in Gemini CLI
$ gemini mcp add -s user data-cleaner-agent uvx -- agentic-csv-cleanerInstall data-cleaner-agent MCP in Gemini settings.json
~/.gemini/settings.json
{
"mcpServers": {
"data-cleaner-agent": {
"command": "uvx",
"args": [
"agentic-csv-cleaner"
]
}
}
}Automated checks
Ran Nov 15, 2026Full report →Passed our automated checks. That is not a security audit: read the code and give the server the narrowest access it needs. How checks work
- Source repository linked and reachablewesseltl/data-cleaner-agent is public and reachablegithub.com/wesseltl/data-cleaner-agent ↗Pass
- Package names this repository as its sourceagentic-csv-cleaner names wesseltl/data-cleaner-agent as sourcepypi.org/project/agentic-csv-cleaner/ ↗Pass
- Build provenance (Sigstore attestation)No build attestation published for this releaseNot checked
- Open source license (OSI)MITspdx.org/licenses/MIT.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-08-10github.com/wesseltl/data-cleaner-agent ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNo popular package one edit away from agentic-csv-cleanerPass
- No critical advisory in the package or its dependencies5 resolved dependencies checked; no critical advisoryosv.dev ↗Pass
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubdata-cleaner-agent
Clean a messy CSV with an agentic workflow. An LLM decides which cleaning steps the data needs, and tested Python functions do the actual work. The model plans the cleanup, it never touches your data values, so nothing gets hallucinated or silently rewritten.
Works offline out of the box (no API key). Can be driven by an LLM, and other AI agents can call it as an MCP tool.
Before / after
Full Name , Country, Signup Date, Amount Paid full_name country signup_date amount_paid
Alice ,Netherlands,2023-01-05,"€1.200,50" Alice Netherlands 2023-01-05 1200.5
Bob,nederland,05/01/2023,"$900" ─────▶ Bob Netherlands 2023-01-05 900.0
Alice ,NL,2023-01-05,"€1.200,50" Carol Germany 2023-02-10 1000.0
Carol , Germany ,2023-02-10,1000 Dan Belgium 2023-03-01 750.0
Dan,belgie,2023/03/01,"€ 750,00"
In one pass it fixed the headers, trimmed whitespace, parsed three different date formats to ISO,
turned €1.200,50 / $900 / € 750,00 into numbers, standardized the country names, and dropped the
duplicate Alice row.
Install & run
pip install agentic-csv-cleaner
clean-csv messy.csv cleaned.csv # clean a file
clean-csv messy.csv # or just print the result
No API key needed. The default planner is a set of offline heuristics.
The idea
An "agentic workflow" is just software with a few parts:
look at the data -> planner picks the steps -> run the steps -> report
(rules, or an LLM) (tested code)
The decision that makes it safe to trust:
The planner decides which tool runs on which column. The tools do the transformation. A language model is good at judgment ("this column looks like money") and bad at being a reliable calculator. So the LLM only ever picks operations from a fixed set. It never reads a value and writes back a "cleaned" one, which is where LLM data-cleaning usually goes wrong.
Two planners, one loop
| Planner | What it is | Needs |
|---|---|---|
RuleBasedPlanner |
Offline heuristics from a quick data profile. The default. | nothing |
LLMPlanner |
Sends the profile + tool list to an LLM, gets back a JSON plan. | pip install "agentic-csv-cleaner[llm]" + ANTHROPIC_API_KEY |
Both return the same list of steps, so the loop is identical. You swap the brain, not the plumbing.
The log reports what each step actually did, including where a conversion could not produce a clean result (unparseable numbers/dates, unmapped categories), so a clean parse is distinguishable from a confident guess.
Use it from other code or agents
from cleaner.api import clean_csv_text
result = clean_csv_text(open("messy.csv").read())
print(result["cleaned_csv"])
print(result["steps"])
As an MCP tool (Claude Desktop)
Other AI agents can call the cleaner as a tool, so they clean a CSV properly instead of reformatting it token by token in the prompt. Three steps:
1. Install it
pip install "agentic-csv-cleaner[mcp]"
2. Add it to your client's config (Claude Desktop's config lives at
~/Library/Application Support/Claude/claude_desktop_config.json on macOS, or
%APPDATA%\Claude\claude_desktop_config.json on Windows):
{
"mcpServers": {
"csv-cleaner": { "command": "python", "args": ["-m", "cleaner.mcp_server"] }
}
}
3. Restart Claude Desktop. The agent now has a clean_csv tool that takes CSV text and returns the
cleaned CSV plus a report of what it did.
Use it with other MCP clients
The same server works in any MCP client, only the config differs. The command is
python -m cleaner.mcp_server.
Cursor — ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project), hot-reloads:
{ "mcpServers": { "csv-cleaner": { "command": "python", "args": ["-m", "cleaner.mcp_server"] } } }
VS Code / GitHub Copilot — .vscode/mcp.json. Note the different key (servers, not mcpServers)
and the required type. Tools only run in Copilot Agent mode:
{ "servers": { "csv-cleaner": { "type": "stdio", "command": "python", "args": ["-m", "cleaner.mcp_server"] } } }
Windsurf — ~/.codeium/windsurf/mcp_config.json (create it if missing):
{ "mcpServers": { "csv-cleaner": { "command": "python", "args": ["-m", "cleaner.mcp_server"] } } }
Cline — add it from the extension's MCP settings panel in VS Code.
Understanding the report
The tool doesn't just hand back tidy data, it tells you what each step actually did, including where it couldn't get a clean result, so you can tell a clean parse from a confident guess:
- coerce_numeric(amount): all 4 value(s) parsed cleanly
- standardize_dates(signup): 1/3 value(s) could not be parsed, set to null
- standardize_categorical(country): 1 value(s) not in the mapping, left unchanged: ['MARS']
So instead of silently dropping a value or leaving a wrong category, it surfaces it, and you know exactly which cells to double-check.
What's in the box
cleaner/tools.py holds the transformations: snake_case_headers, strip_whitespace,
coerce_numeric, standardize_dates, standardize_categorical, drop_duplicate_rows.
Take standardize_dates. 2023-01-05 (year first, month in the middle) and 05/01/2023 (day first)
need opposite parsing rules, and one global setting corrupts one or the other, so the tool decides per
value. Mixed dates are genuinely ambiguous, and this handles them explicitly instead of guessing.
Tests
python -m unittest discover -s tests
License
MIT.
data-cleaner-agent MCP questions
How do I add data-cleaner-agent MCP to Claude Code?
Run claude mcp add data-cleaner-agent -- uvx agentic-csv-cleaner in your terminal. Add --scope user to make it available in every project.
How do I add data-cleaner-agent MCP to Cursor?
Open ~/.cursor/mcp.json and add the server under mcpServers, using the Cursor config shown in the install section above.
How can I tell if a data conversion was successful?
The tool provides a report detailing what each step did, specifically highlighting where conversions could not produce a clean result.