# letin: Power Query (M) (pypi · letin)

Inspect, refactor, lint, run and version Power Query (M) in Excel workbooks and Power BI projects.

- Trust score: 49/100 (low)
- Registry status: active
- Liveness: live
- Owner verified: no
- Last scored: 2026-09-22

## Components

- pypi · `letin`: 49/100 (this document), [markdown](https://verifymcp.io/servers/s-curvelabs-letin/letin.md), [page](https://verifymcp.io/servers/s-curvelabs-letin/letin)

## Channel facts

- Registry: `pypi`
- Package: `letin`
- Version: `0.2.0`
- Transport: `stdio`

## Trust breakdown

How this component scores in each security and reliability category. Every signal is checked automatically from public evidence about the published package, including repeated runs of it in an isolated sandbox, and we only credit what we can confirm. Scores are 0–100 per category. Scoring method: https://verifymcp.io/docs/scoring (what has changed: https://verifymcp.io/docs/scoring/changelog)

Scored 2026-09-22.

- **Supply Chain Security**: 50/100
  - Malware scan not yet available for this package.
  - No known CVEs affecting this package version or its production dependencies.
  - Runs hatchling.build at install time, a recognised native-build step with no shell scripting around it.
  - 1 of 29 dependencies flagged as unhealthy.
- **Provenance & Transparency**: 32/100
  - Source repository is publicly reachable at the declared URL.
  - Provenance check failed: no build-provenance attestation is published.
  - License check failed: no license is declared.
  - Actively maintained (last published 0 days ago).
  - Disclosure check failed: no security disclosure policy was found in the source repository.
- **Schema Quality & AI Usability**: 73/100
  - AI-judged instruction clarity (good).
  - Tool/resource definitions use about 2726 tokens (~87/item across 31 items; 31 tools + 0 resources), lean.
  - Usage-examples check failed: none of the tools include examples.
- **Stability & Change Management**: 0/100
  - Stability not yet verified: not enough scan history yet (needs a 30-day window).
- **Tool Coverage**: 67/100
  - 100% of tools have a non-trivial description (not blank, and not just the tool's name).
  - 0% of tool parameters carry a description.
- **Tool Safety**: 88/100
  - No prompt-injection markers were found in the server instructions, tool names or descriptions we captured.
  - 1 of 2 tool(s) whose name or description implies an irreversible operation declare an MCP destructiveHint annotation; "remove_step" implies "remove" and declares no destructiveHint at all, which the MCP spec reads as destructive by default.
  - An AI judge read all 32 captured unit(s) of tool text and found none that tries to manipulate the model reading it.
- **Capabilities**: 100/100
  - Implements a current MCP spec version (2026-07-28).

**Unverified: 1 category.** A category scored 0 because we could not verify it: a data source with nothing on this package, evidence we could not reach, or a check we could not run. We only credit what we can confirm.

## Install

### How do I install the letin: Power Query (M) MCP server?

letin: Power Query (M) runs locally as a PyPI package, launched with uvx letin. Ready-made configuration for Claude, Cursor, VS Code, Codex and 5 more is on this page, copied from each client's own documentation.

### Claude

```bash
claude mcp add s-curvelabs-letin -- uvx letin
```

### Cursor

```json
{
  "mcpServers": {
    "s-curvelabs-letin": {
      "command": "uvx",
      "args": [
        "letin"
      ]
    }
  }
}
```

### VS Code

```json
{
  "servers": {
    "s-curvelabs-letin": {
      "command": "uvx",
      "args": [
        "letin"
      ]
    }
  }
}
```

### Codex

```bash
codex mcp add s-curvelabs-letin -- uvx letin
```

### opencode

```json
{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "s-curvelabs-letin": {
      "type": "local",
      "command": [
        "uvx",
        "letin"
      ],
      "enabled": true
    }
  }
}
```

### OpenClaw

```bash
openclaw mcp add s-curvelabs-letin --command uvx --arg letin
```

### Hermes

```yaml
mcp_servers:
  s-curvelabs-letin:
    command: "uvx"
    args: ["letin"]
```

### Netclaw

```json
{
  "McpServers": {
    "s-curvelabs-letin": {
      "Transport": "stdio",
      "Command": "uvx",
      "Arguments": [
        "letin"
      ]
    }
  }
}
```

### Vellum

```bash
assistant mcp add s-curvelabs-letin -t stdio -c uvx -a letin
```

### Other

```json
{
  "mcpServers": {
    "s-curvelabs-letin": {
      "command": "uvx",
      "args": [
        "letin"
      ]
    }
  }
}
```

## Changelog

Every change recorded for this component, newest first. Days that predate change tracking, or that we cannot explain, say so: "we were watching and nothing happened" and "we were not watching" are different claims.

### 2026-09-22 (score 49)

First indexed and scored.

## MCP tools (31)

### `list_sources` (~36 tokens)

List workbooks open in Excel and models open in Power BI Desktop (both usable as `source`), and explain the other source forms.

### `list_queries` (~42 tokens)

Every query with its kind (query/function/parameter), step count, load destination, and what it depends on / is used by.

Input parameters:

- `source` (string, required)

### `get_query` (~41 tokens)

Full M code of one query plus its parsed steps, data sources, dependencies and dependents.

Input parameters:

- `name` (string, required)
- `source` (string, required)

### `dependency_graph` (~71 tokens)

Query-to-query lineage. With `query`, only its upstream and downstream. format: 'json' or 'mermaid'.
Also reports roots (read external data), leaves (final outputs), orphans and cyclic references.

Input parameters:

- `format` (string)
- `query`
- `source` (string, required)

### `list_data_sources` (~44 tokens)

Inventory of every external source (files, folders, servers, URLs) each query reads, and whether it is hard-coded or parameterised.

Input parameters:

- `source` (string, required)

### `find_in_queries` (~64 tokens)

Search the M of every query. scope: all | code | strings | comments. Returns query, step and line for each hit.

Input parameters:

- `pattern` (string, required)
- `regex` (boolean)
- `scope` (string)
- `source` (string, required)

### `lint_queries` (~78 tokens)

Audit queries for folding breakers, hard-coded paths/servers, dead steps, brittle column lists, untyped columns,
duplicated logic, orphans and cycles. Each finding carries a concrete suggestion. min_severity: info | warning | error.

Input parameters:

- `min_severity` (string)
- `query`
- `source` (string, required)

### `set_query` (~94 tokens)

Replace a query's M (or create it with create_if_missing). Rejects unbalanced brackets/strings before writing.
For standard actions prefer apply_transform; for renames prefer rename_query / rename_step.

Input parameters:

- `create_if_missing` (boolean)
- `description` (string)
- `dry_run` (boolean)
- `formula` (string, required)
- `name` (string, required)
- `source` (string, required)

### `delete_query` (~52 tokens)

Delete a query. Refuses when other queries reference it unless force=true.

Input parameters:

- `dry_run` (boolean)
- `force` (boolean)
- `name` (string, required)
- `source` (string, required)

### `rename_query` (~69 tokens)

Rename a query and rewrite every reference to it in all other queries (token-aware: strings, comments and
same-named columns are left alone).

Input parameters:

- `dry_run` (boolean)
- `new_name` (string, required)
- `old_name` (string, required)
- `source` (string, required)

### `rename_step` (~58 tokens)

Rename a step inside a query and every reference to it.

Input parameters:

- `dry_run` (boolean)
- `new_name` (string, required)
- `old_name` (string, required)
- `query` (string, required)
- `source` (string, required)

### `remove_step` (~55 tokens)

Delete a step and rewire the steps that used it to the step it read from.

Input parameters:

- `dry_run` (boolean)
- `query` (string, required)
- `source` (string, required)
- `step` (string, required)

### `replace_in_queries` (~118 tokens)

Find and replace across all (or the listed) queries in one go, e.g. repoint a server or folder.
scope: 'strings' (only inside text literals; safest, the default), 'code' (outside strings and comments), 'all'.

Input parameters:

- `dry_run` (boolean)
- `find` (string, required)
- `ignore_case` (boolean)
- `queries`
- `regex` (boolean)
- `replace` (string, required)
- `scope` (string)
- `source` (string, required)

### `extract_parameter` (~102 tokens)

Turn a hard-coded text value (path, server, URL) into a Power Query parameter and reference it everywhere.
With match_prefix, "C:\Data" also rewrites "C:\Data\sales.csv" to SourceFolder & "\sales.csv".

Input parameters:

- `dry_run` (boolean)
- `match_prefix` (boolean)
- `parameter_name` (string, required)
- `queries`
- `source` (string, required)
- `value` (string, required)

### `list_transforms` (~36 tokens)

Catalog of guided actions for apply_transform (unpivot, pivot, group_by, merge, filter, types, ...) with their params.

### `apply_transform` (~164 tokens)

Add a standard Power Query step without hand-writing M. `action` and `params` come from list_transforms, e.g.
action='unpivot_other_columns', params={'keep_columns': ['Region']}; or action='group_by',
params={'group_columns': ['Region'], 'aggregations': [{'name': 'Total', 'operation': 'sum', 'column': 'Amount'}]}.
The step is appended to the end unless after_step names an earlier step (the next step is rewired, as the editor does).

Input parameters:

- `action` (string, required)
- `after_step`
- `dry_run` (boolean)
- `params` (object, required)
- `query` (string, required)
- `source` (string, required)
- `step_name`

### `copy_queries` (~72 tokens)

Copy queries between workbooks / projects / folders, bringing the queries they depend on along.

Input parameters:

- `dry_run` (boolean)
- `from_source` (string, required)
- `include_dependencies` (boolean)
- `names` (array, required)
- `overwrite` (boolean)
- `to_source` (string, required)

### `export_queries` (~62 tokens)

Write every query to `folder` as one .pq file each plus pq-manifest.json: ready to commit to git and diff in a PR.
Existing .pq files in the folder are replaced.

Input parameters:

- `folder` (string, required)
- `source` (string, required)

### `import_queries` (~74 tokens)

Load .pq files from `folder` back into `source`. mode='update' changes/adds queries; mode='sync' also deletes
queries that are not in the folder.

Input parameters:

- `dry_run` (boolean)
- `folder` (string, required)
- `mode` (string)
- `source` (string, required)

### `diff_queries` (~68 tokens)

Compare the queries of two sources (workbook vs workbook, workbook vs exported folder, PBIP vs PBIP ...).
Line endings are ignored. To compare against a snapshot pass its `path` from list_snapshots.

Input parameters:

- `source_a` (string, required)
- `source_b` (string, required)

### `list_snapshots` (~30 tokens)

Snapshots taken automatically before each write to this source, newest first.

Input parameters:

- `source` (string, required)

### `restore_snapshot` (~77 tokens)

Roll back to a snapshot. With `queries`, only those are restored; otherwise the whole query set is put back
(queries created since are deleted). 'latest' undoes the most recent write.

Input parameters:

- `dry_run` (boolean)
- `queries`
- `snapshot_id` (string)
- `source` (string, required)

### `validate_m` (~119 tokens)

Check M before using it: syntax errors with line/column and a plain-English cause (missing comma between steps,
unquoted step name, if without else ...), unknown steps/queries/functions with did-you-mean, and wrong argument counts
for library functions. Pass `formula` (an expression or a whole section document); add `source` so its query names are
known. Or pass `source` (+ optional `query`) alone to validate what is already there.

Input parameters:

- `formula`
- `query`
- `source`

### `format_m` (~58 tokens)

Pretty-print M (an expression or a section document): one step per line, long calls broken at their arguments,
comments kept. The output is verified token-for-token against the input, so meaning can never change.

Input parameters:

- `formula` (string, required)

### `format_queries` (~45 tokens)

Format one query, or every query, in place (same guarantees as format_m).

Input parameters:

- `dry_run` (boolean)
- `query`
- `source` (string, required)

### `m_function_help` (~96 tokens)

Look up the M standard library (generated from the real engine: 850+ members). Give an exact name
('Table.Group') for its signature and description, a namespace ('Table.' or 'List') to list its members, or words to
search names and descriptions ('unpivot', 'split text delimiter'). Use it instead of guessing signatures.

Input parameters:

- `limit` (integer)
- `name_or_search` (string, required)

### `evaluate_m` (~199 tokens)

Run any M expression and get the result as data. Tables come back with column names, types and the first `rows`
rows (cell errors shown as #ERROR); lists, records, scalars and functions (with signature) are described too; an M error
comes back as reason + message. With `source`, the expression can reference that source's queries by name
(e.g. Table.RowCount(Sales)), including unsaved edits of an open workbook. count_rows and profile (Table.Profile:
min/max/nulls/distinct per column) scan the whole result, so they can be slow. ignore_privacy_levels=true disables the
privacy firewall for this throwaway evaluation only.

Input parameters:

- `count_rows` (boolean)
- `expression` (string, required)
- `ignore_privacy_levels` (boolean)
- `profile` (boolean)
- `rows` (integer)
- `source`
- `timeout_seconds` (integer)

### `evaluate_query` (~131 tokens)

Freshly evaluate a query, or the query only up to `step`, and return columns, types and sample rows. Use it to get real
column names before apply_transform, to see what a step does to the data, or to find the step where an error first
appears (evaluate step by step).

Input parameters:

- `count_rows` (boolean)
- `ignore_privacy_levels` (boolean)
- `profile` (boolean)
- `query` (string, required)
- `rows` (integer)
- `source` (string, required)
- `step`
- `timeout_seconds` (integer)

### `evaluate_steps` (~109 tokens)

Evaluate EVERY step of a query in one engine call and describe each one (columns, types, first `rows` rows, or the
error). The fastest way to see how a query transforms its data and to find the step where an error first appears:
one call instead of evaluate_query once per step.

Input parameters:

- `ignore_privacy_levels` (boolean)
- `query` (string, required)
- `rows` (integer)
- `source` (string, required)
- `timeout_seconds` (integer)

### `refresh_query` (~47 tokens)

Refresh one query in an open Excel workbook and wait for it; returns the mashup engine's error text if it fails.

Input parameters:

- `name` (string, required)
- `source` (string, required)

### `preview_query` (~61 tokens)

Column names and first rows of a query that is loaded to a worksheet table (open Excel workbook only).
Use it to get real column names before apply_transform.

Input parameters:

- `name` (string, required)
- `rows` (integer)
- `source` (string, required)

## Diagnostics

Captured diagnostic sections: Provenance, Install scripts, Dependencies. The full working is on the page: https://verifymcp.io/servers/s-curvelabs-letin/letin#diagnostics

## Score history

- 2026-09-22: 49

## Common questions

### What is the letin: Power Query (M) MCP server?

letin: Power Query (M) is an MCP server listed in the public MCP registry as io.github.S-CurveLabs/letin. Inspect, refactor, lint, run and version Power Query (M) in Excel workbooks and Power BI projects. This page covers its PyPI package (letin).

### Is the letin: Power Query (M) MCP server safe to use?

letin: Power Query (M) scores 49 out of 100 on VerifyMCP. We found no known CVEs affecting it as of 22 September 2026. That is a record of what we were able to check automatically, not an endorsement. The category breakdown on this page shows every signal behind the number, including the ones we could not confirm.

### What tools does the letin: Power Query (M) MCP server expose?

letin: Power Query (M) exposes 31 tools: list_sources, list_queries, get_query, dependency_graph, list_data_sources, and 26 more. Their descriptions and schemas cost roughly 2,372 tokens of context every time the server is loaded.

### Is the letin: Power Query (M) MCP server still maintained?

letin: Power Query (M) is still listed as active in the MCP registry. We last reached this channel on 22 September 2026. Those dates come from our own scans of the registry and the channel itself, not from anything the publisher announced.

## Links

- PyPI project: https://pypi.org/project/letin/
- Socket report: https://socket.dev/pypi/package/letin
- Repository: https://github.com/S-CurveLabs/letin
- Changelog RSS feed: https://verifymcp.io/servers/s-curvelabs-letin/letin.xml
- Changelog JSON feed: https://verifymcp.io/servers/s-curvelabs-letin/letin.json
- HTML version of this page: https://verifymcp.io/servers/s-curvelabs-letin/letin