MCP server for Japanese labor and social insurance law — e-Gov API, MHLW rulings, JAISH safety and health rulings
DevRay DataLab is installed from its publisher's own source and answers where it runs, so this marketplace is not in the path of a single call. There is no address here to send one to, and a panel that pretended otherwise would be showing you an answer we made up. Install it and call it from your own client — the Installation tab has the entry for each one.
What it does
Local-first data science and machine learning MCP server
Quickstart
# 1 — add it to Claude Code
claude mcp add --scope user devray-datalab-mcp '--' uvx devray-datalab-mcp
# 2 — ask your agent for something these tools do
# list_datasets, list_split_manifests, create_split_manifest and 24 more
DevRay DataLab costs nothing on mcprush: there is no plan to choose here, no cap to set and nothing mcprush can bill you for.
Where are you running it?
Every route below installs the same thing and ends at the same approval screen. This one runs on your machine: your client starts DevRay DataLab as a process under your own user, with your files and your network, so the tool surface below is what it can do to your computer rather than to a server somewhere else. Its scan result is on the Security tab — read the surface before you approve it.
This is a public server: you run it yourself and this marketplace is not in the path. You need uv (docs.astral.sh/uv): uvx comes with it. Claude Code registers it in one command; --scope user makes it available in every project. The '--' is in quotes so that PowerShell passes it on.
claude mcp add --scope user devray-datalab-mcp '--' uvx devray-datalab-mcpReconnect, or start a new session, and the tools appear in the model’s tool list.
One config entry your client uses to start the process locally. A local server runs with your file system and your network, which is why it is priced without metering.
27 tools, with what each one reads, writes and reaches shown before you agree — the same list on every route above. Read the tool surface.
Tool surface
What the model actually sees. Descriptions are diffed on every release — see version history.
Read from its source code. A server that builds its tool list when it starts can list more tools than this.
Discover dataset filenames, formats and sizes in the configured workspace without reading rows. Lists CSV, Parquet and Excel files; use inspect_dataset to profile one.
No parameter schema on file.
Find stored split manifest IDs and metadata, newest first. Use validate_split_manifest to check an ID against the current dataset.
No parameter schema on file.
Save reproducible row indices and a dataset fingerprint before training. Pass the returned ID as split_manifest_id to compatible training tools; keep test rows out of tuning. Use validate_split_manifest to check a saved split without rewriting it.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| task_type | string | no | 'classification' or 'regression' (default classification). |
| split_type | string | no | random (default), stratified (pass stratify_column), group (requires group_column), or time (requires time_column or row_order=true). grouped and time_series are aliases. |
| test_size | number | no | Fraction of data for test holdout (default 0.2). |
| val_size | number | no | Validation fraction (default 0); test_size + val_size < 1. |
| random_state | integer | no | Random seed for reproducibility (default 42). |
| stratify_column | string | no | Column to stratify splits by. |
| group_column | string | no | Column to group splits by (groups stay together). |
| time_column | string | no | Column for time-based splitting. |
| row_order | boolean | no | Split chronologically by original row order. |
| protected_holdout | boolean | no | Keep test set as protected holdout (default true). |
| delimiter | string | no | CSV delimiter (comma, semi, tab, pipe). Auto. |
Check a saved manifest against the named workspace dataset before reuse. Returns valid and any errors for fingerprint or split mismatches; does not repair the file. Find IDs with list_split_manifests.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| manifest_id | string | yes | ID of the split manifest to validate. |
| name | string | yes | Dataset filename. |
Profile a named workspace dataset: shape, column types, missingness, summaries and a bounded preview that includes raw values. Use analyze_data_quality for remediation recommendations. CSV is built in; Parquet and Excel readers need optional packages.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| max_rows | integer | no | Max preview rows (default 25, max 100). |
Rank numeric column pairs by correlation strength within one dataset. Use Pearson for linear association or Spearman for rank association; neither establishes causation. For changes between datasets, use analyze_feature_drift.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| method | string | no | 'pearson' or 'spearman' (default pearson). |
| top_n | integer | no | Number of top pairs to return (default 10). |
Check a workspace dataset for missing data, duplicates, constants and other quality issues before modeling. Returns recommendations without modifying the dataset. Use inspect_dataset for a raw row preview.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
Screen a workspace dataset before fitting for potential leakage, such as ID-like columns and, when target_column is supplied, predictors suspiciously related to the label. These flags need domain review; the dataset is not modified.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| target_column | string | no | Target variable column (optional). |
| delimiter | string | no | CSV delimiter (comma, semi, tab, pipe). Auto. |
Compare feature distributions in a reference (baseline) and current workspace dataset, using shared columns unless a subset is supplied. Reports statistical drift flags at alpha without modifying data; use analyze_data_quality for one dataset.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| reference_name | string | yes | Reference dataset filename. |
| current_name | string | yes | Current dataset filename. |
| alpha | number | no | Significance level (default 0.05). |
| columns | array | no | Columns to check (default all common). |
| ref_delimiter | string | no | Reference CSV delimiter (comma, semi, tab, pipe). |
| cur_delimiter | string | no | Current CSV delimiter (comma, semi, tab, pipe). |
Split a named dataset for classification or regression and fit preprocessing on training rows only. Saves a preprocessing artifact and returns its handle and schema. Training tools fit their own preprocessing.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| target_column | string | yes | Target variable column. |
| task_type | string | yes | 'classification' or 'regression'. |
| test_size | number | no | Test fraction (default 0.2). |
| random_state | integer | no | Random seed (default 42). |
| stratify | boolean | no | Stratify split (classification only, default true). |
Fit one named model and return evaluation metrics and feature metadata. Supply a dataset, target and task. Use start_training_job for background evaluation or export_model to fit and save a reusable bundle.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| target_column | string | yes | Target variable column. |
| task_type | string | yes | 'classification' or 'regression'. |
| model_name | string | yes | Model to train. |
| test_size | number | no | Test fraction (default 0.2). |
| random_state | integer | no | Random seed (default 42). |
| stratify | boolean | no | Stratify split (classification only, default true). |
Fit and compare candidate algorithms on a dataset using cross-validation (CV). Returns fold metrics and selects a winner by the requested metric. Provide a split_manifest_id to reserve its test rows. Records experiments when possible. Use compare_experiments for saved metrics or automl_run for…
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| target_column | string | yes | Target variable column. |
| task_type | string | yes | 'classification' or 'regression'. |
| models | array | no | Models to compare. |
| metric | string | no | Selection metric (default auto). |
| test_size | number | no | Test fraction (default 0.2). |
| random_state | integer | no | Random seed (default 42). |
| cv | integer | no | Cross-validation folds (default 5). |
Write a named local model bundle with fitted preprocessing and feature schema for predict. Supply name, target_column, task_type and model_name to train, or pass a trusted existing bundle name as model_handle to copy it. Reusing artifact_name can overwrite a saved bundle.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| artifact_name | string | yes | Name for the saved artifact. |
| model_handle | string | no | Handle of a previously trained model. |
| name | string | no | Dataset filename (to train and export). |
| target_column | string | no | Target variable column. |
| task_type | string | no | 'classification' or 'regression'. |
| model_name | string | no | Model to train and export. |
| test_size | number | no | Test fraction (default 0.2). |
| random_state | integer | no | Random seed (default 42). |
| stratify | boolean | no | Stratify split (classification only). |
Score a feature object or array of feature objects with a trusted local model bundle, without fitting. Match feature keys to the stored schema; use load_model_handle to inspect it. Never load artifacts from untrusted sources.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| artifact_name | string | yes | Name of the exported model artifact. |
| features | yes | Feature record(s) to predict on. |
Check the server's CPU, RAM, storage and detected GPUs before a costly run. This does not train or enable GPU acceleration.
No parameter schema on file.
Browse recent persisted experiment records, newest first, to find IDs for get_experiment or compare_experiments. Use list_training_jobs for in-process jobs.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| limit | integer | no | Max experiments to return (default 50). |
Read one saved experiment's metrics and metadata by ID from list_experiments. Use compare_experiments for several runs; this does not fit or load a model.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| experiment_id | string | yes | Experiment ID. |
Compare saved metrics for experiment IDs from list_experiments without retraining. Interpret scores only across comparable datasets, splits and metrics; use compare_models to run a new comparison on one dataset.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| experiment_ids | array | yes | List of experiment IDs. |
Find experiment records that reference model artifacts and their identifiers. Use load_model_handle to confirm a bundle is available and read its schema.
No parameter schema on file.
Inspect feature schema, metrics and metadata from a trusted saved model bundle. Supply an artifact_name or an experiment_id with a tracked artifact. This loads the bundle but does not score rows; use predict for inference.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| experiment_id | string | no | Experiment ID. |
| artifact_name | string | no | Model artifact name. |
Queue one named model fit and return a job_id without waiting for training. Supply a dataset, target and task, then poll get_training_status for results. Use train_baseline when you want the metrics in the same call.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| target_column | string | yes | Target variable column. |
| task_type | string | yes | 'classification' or 'regression'. |
| model_name | string | yes | Model to train. |
| test_size | number | no | Test fraction (default 0.2). |
| random_state | integer | no | Random seed (default 42). |
| stratify | boolean | no | Stratify split (classification only, default true). |
| exclude_columns | array | no | Feature columns to exclude before preprocessing/training. |
| split_manifest_id | string | no | ID of a validated split manifest to enforce indices. |
| delimiter | string | no | CSV delimiter (comma, semicolon, tab, pipe). Auto. |
Poll a job_id from start_training_job or list_training_jobs for state and any available result or error. Job IDs are process-local.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| job_id | string | yes | Training job id. |
Request cancellation of a queued or running job by job_id. Running work stops cooperatively, so poll get_training_status to confirm the terminal state. A request does not undo finished work.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| job_id | string | yes | Training job id. |
Discover recent jobs in this server process, newest first. Use get_training_status for one job's result or list_experiments for persistent run history.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| limit | integer | no | Max jobs to return (default 50). |
Run a fresh, synchronous cross-validation search and return ranked trials, winner and run metadata. Optionally tune the leading candidate when Optuna and budget remain. Provide a split_manifest_id to reserve its test rows. Use tune_model for one algorithm. Search returns scores and parameters…
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| target_column | string | yes | Target variable column. |
| task_type | string | yes | 'classification' or 'regression'. |
| metric | no | — | |
| max_trials | integer | no | Total candidate and tuning trials, 1–50 (default 10). |
| max_time_seconds | no | — | |
| random_state | integer | no | Random seed (default 42). |
| cv | integer | no | Cross-validation folds (default 3). |
| models | no | — | |
| exclude_columns | array | no | Feature columns to exclude before preprocessing/training. |
| split_manifest_id | string | no | ID of a validated split manifest to enforce indices. |
| delimiter | no | — |
Score one model by cross-validation, then search its hyperparameters if Optuna is installed and budget remains. Without Optuna, only the baseline is scored. Returns scores and parameters. Use automl_run to search across model families.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| target_column | string | yes | Target variable column. |
| task_type | string | yes | 'classification' or 'regression'. |
| model_name | string | yes | Model to tune. |
| metric | no | — | |
| max_trials | integer | no | Total trials including the baseline, 1–50 (default 20). |
| max_time_seconds | no | — | |
| random_state | integer | no | Random seed (default 42). |
| cv | integer | no | Cross-validation folds (default 3). |
| exclude_columns | array | no | Feature columns to exclude before preprocessing/training. |
| split_manifest_id | string | no | ID of a validated split manifest to enforce indices. |
| delimiter | no | — |
Run fresh cross-validation and return ranked trials plus the winner's model ID, parameters and score. Pass models to restrict candidates. Use automl_run for full run metadata or list_experiments to find saved training records. Use export_model to fit and save a model bundle.
| Parameter | Type | Required | Notes |
|---|---|---|---|
| name | string | yes | Dataset filename. |
| target_column | string | yes | Target variable column. |
| task_type | string | yes | 'classification' or 'regression'. |
| metric | no | — | |
| max_trials | integer | no | Total candidate and tuning trials, 1–50 (default 10). |
| max_time_seconds | no | — | |
| random_state | integer | no | Random seed (default 42). |
| cv | integer | no | Cross-validation folds (default 3). |
| models | no | — | |
| exclude_columns | array | no | Feature columns to exclude before preprocessing/training. |
| split_manifest_id | string | no | ID of a validated split manifest to enforce indices. |
| delimiter | no | — |
- Every tool, no call limit
- No card, and mcprush charges nothing for it
- Source published under MIT
- Runs on your machine — nothing of it reaches our gateway
- Nothing to cap, because mcprush bills nothing
What counts against your monthly calls
| Tool | Unit | Calls used | Out of the allowance |
|---|
Nothing here is billable: mcprush charges nothing to install DevRay DataLab or to call it, at any volume.
Two independent axes, because powerful and malicious are different questions. The grade is threat only. The capability level is blast radius, and it is never a penalty on the grade — it is priced as one subtract-only term in the score, where you can see it.
| Term | Level | What it prices | Points |
|---|---|---|---|
| capability-exposure | minimal | capability blast radius (minimal) — client exposure if the model is manipulated | −0 |
| verification-discount | source | publisher verification (provenance) — cryptographic build provenance ties the artifact to its source | −0 |
| coverage-honesty | source | inspection depth (source) — how much of the target the scan could see | −0 |
What the scan could actually read
A grade is only as meaningful as its coverage, so the scanner publishes its own depth before it publishes its result.
Tools were statically extracted from the published source (27 recovered), not enumerated from a running server. Tool-poisoning, Unicode-smuggling, capability and toxic-flow analysis ran on this inferred surface, but a mis-parsed registration could be missed or mis-attributed, so tool-derived findings are capped below “confirmed”. To grade the real runtime surface, scan the running server: --command "npx -y <package>".
Capability — what it could do if the model were manipulated
Tags derived from each tool’s schema and the implementation, not from what the tool calls itself. minimal is the level these add up to.
| Tool | Capability tags | Why the tag was assigned |
|---|---|---|
| list_datasets | no tags | |
| list_split_manifests | no tags | |
| create_split_manifest | no tags | |
| validate_split_manifest | no tags | |
| inspect_dataset | no tags | |
| analyze_correlations | no tags | |
| analyze_data_quality | no tags | |
| detect_potential_leakage | no tags | |
| analyze_feature_drift | no tags | |
| prepare_dataset | no tags | |
| train_baseline | no tags | |
| compare_models | no tags | |
| export_model | no tags | |
| predict | no tags | |
| system_info | no tags | |
| list_experiments | untrusted-input | untrusted-input: keyword "browse" |
| get_experiment | no tags | |
| compare_experiments | no tags | |
| list_models | no tags | |
| load_model_handle | no tags | |
| start_training_job | no tags | |
| get_training_status | no tags | |
| cancel_training_job | no tags | |
| list_training_jobs | no tags | |
| automl_run | no tags | |
| tune_model | no tags | |
| get_model_leaderboard | no tags |
Toxic-flow graph
The lethal trifecta, checked as a graph rather than as a checklist: untrusted input, a sensitive source and an external sink have to meet before there is a path worth worrying about.
The public result for this release does not print the flow graph, so there is nothing to show here. That is not the same as "no paths were found": what the scan did read is above, under coverage.
Supply chain and provenance
This is the first scan of this surface here, so there is nothing yet to compare it against.
Every result on this tab comes from one deterministic pass over the published package — offline, rule by rule, and auditable line by line above. Same methodology version, same bytes, same score.
Release history
Always latest — the install command names no version, so it installs the newest release when it runs; the current release on file is 0.3.4. Pin a release below and the command asks for exactly that one.
- DevRay DataLab MCP v0.3.4 improves reliability, model selection, and integration with AI assistants. Adds optional model filtering in `get_model_leaderboard`, improves descriptions for all 27 MCP tools, supports explicit CSV delimiters in AutoML workflows, and strengthens reproducible tuning, chronological cross-validation, cancellation, and artifact validation. Available on GitHub, PyPI, and the official MCP Registry.
No release note was published with this version.
No release note was published with this version.
No release note was published with this version.
- The version this listing was on when mcprush began following its releases. The registry was not asked when it shipped, so this row carries no date.
One review per account, from a signed-in account that does not publish this listing. Nothing else is asked — you can say what you think before you install it. Publishers can reply once per review, and a review can be edited or deleted by whoever wrote it.
Any signed-in account that does not publish DevRay DataLab can review it, once — before installing it or after. A review from an account that has installed it is marked as one.
Nobody has reviewed this listing. The rating on the card is the mean of the reviews written here and nothing else, so there is no rating until somebody writes the first — which takes a signed-in account that does not publish it, and nothing else.