# Import Check: Validate & Clean CSV

> Validate and prepare CSV; reconcile two tables with exact keys and full findings. Fixed free demos.

- name: io.github.LeBorgneAntoine/import-check
- version: 0.2.0
- connection class: R0 (autonomous)
- trust: 80/100 flags: no-repository
- quality: 96/100 (strong)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/io.github.LeBorgneAntoine%2Fimport-check
- tools: https://api.protogrid.dev/v1/servers/io.github.LeBorgneAntoine%2Fimport-check/tools

**An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "import-check": {
      "type": "http",
      "url": "https://check.orvel.dev/mcp",
      "headers": {
        "Authorization": "Bearer ${IMPORT_CHECK_TOKEN}"
      }
    }
  }
}
```

Placeholders to fill: `${IMPORT_CHECK_TOKEN}`.

## Trust

- hygiene: 80
- liveness: 100
- freshness: 100
- provenance: 45
- drivers: +website +reachable +uptime-100% +updated-19d-ago -no-repository
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 96/100 (strong)

### protocol: 88 (weight 25)

- pass `protocol.modern`: supports 2026-07-28 (server/discover)
- pass `protocol.stateless`: answers without a session
- pass `protocol.transport`: streamable HTTP
- warn `protocol.list_ttl`: tools/list declares ttlMs 0, so clients re-fetch it on every use

### authorization: n/a (weight 15)

- n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth
- n/a `auth.secret_in_url`: no templated URL

### tool hygiene: 100 (weight 30)

- pass `tools.descriptions`: every tool has a description
- pass `tools.description_length`: descriptions are concise
- pass `tools.schemas`: every tool has an object input schema
- pass `tools.annotations`: 10 of 10 tools declare readOnlyHint or destructiveHint
- pass `tools.directory_hints`: every tool declares readOnlyHint, destructiveHint, idempotentHint and openWorldHint
- pass `tools.token_cost`: about 4,096 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

### stability: 100 (weight 15)

- pass `stability.changes`: 5 tool changes in 30 days
- pass `stability.rug_pull`: no tool changed its meaning under the same name

### dependencies: n/a (weight 15)

- n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package

- tools: 10, about 4,096 tokens to load them all
- tool set hash: 28d059bfa0d732fd254ec39d7045d429
- tools last changed: 2026-10-01T23:00:24.584Z
- badge: [![protogrid quality](https://protogrid.dev/badge/io.github.LeBorgneAntoine/import-check.svg)](https://protogrid.dev/servers/io.github.LeBorgneAntoine/import-check)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

- 2026-10-01 `reconcile_tables` added
  - Join two bounded tables using exact string or composite keys. Both sides must have non-empty unique keys: duplicates block the entire output, including many-to-many matches. Choose left, inner or full retention explicitly; unmatched records are reported even if excluded. Optional compare fields flag changed values without resolving them. Output includes row lineage; missing-side fields are empty strings. No trimming, fuzzy matching or destination writes. Combined inputs share 1 MB/10,000 records/250,000 cells. One completed report costs one operation; reuse request_id for identical retries. Full findings or an uncharged size error, never silent truncation.
- 2026-10-01 `prepare_table` added
  - Select, rename, derive and convert columns using a caller-owned version-1 recipe. Rules reference original input headers; output follows recipe order and omitted columns are listed. All input rows remain. Any failed conversion returns valid=false, all findings and csv_text=null, never a partially transformed file. No arithmetic, arbitrary code, schema inference or destination writes. One completed report costs one operation, including valid=false. Use reconcile_tables for two-file matching.
- 2026-10-01 `inspect_table` added
  - Profile a CSV or explicit string table before writing a recipe. Return empty, distinct and duplicate-value counts per column and complete formula warnings; no schema or business validity is inferred. Supply CSV OR columns/rows, never both. JSON numbers must be encoded as strings so identifiers keep leading zeros. One paid operation; no content retained. Reuse request_id only for identical input.
- 2026-10-01 `view_table_demo` added
  - Join the bundled fixed tables and show changed and unmatched records. No arguments, key or credit. This cannot process custom data; use reconcile_tables with a paid key for your files.
- 2026-10-01 `get_table_example` added
  - Return fixed synthetic left/right tables and a reusable join recipe. No arguments, key or credit. Copy the structure for reconcile_tables; mode controls which unmatched rows appear in output.

Full history: https://api.protogrid.dev/v1/servers/io.github.LeBorgneAntoine%2Fimport-check/changes

## Remotes

- https://check.orvel.dev/mcp (streamable-http, auth none, reachable true, uptime30d 1)

## Packages

_none_

## Tools (10)

- `clean_csv`: Apply only requested column renames and transformations, then validate the output against schema. Cleanup column names refer to the original headers; schema names refer to the renamed output. Transformations run in listed order. decimal_comma changes only plain numbers such as 12,50; ambiguous grouping stays unchanged. No records or columns are removed. Returned csv_text is UTF-8-compatible with CRLF record endings. One paid operation includes cleanup and validation. Spreadsheet-safe escaping is opt-in and may make numeric negative cells fail numeric validation; inspect the final report before import.
- `get_csv_example`: Return a fixed synthetic CSV and its column schema. Free, no key or quota. Copy the structure, then supply your own CSV to validate_csv.
- `get_csv_usage`: Read remaining operations and expiry for the API key supplied in the Authorization header. Does not consume credit.
- `get_table_example`: Return fixed synthetic left/right tables and a reusable join recipe. No arguments, key or credit. Copy the structure for reconcile_tables; mode controls which unmatched rows appear in output.
- `inspect_table`: Profile a CSV or explicit string table before writing a recipe. Return empty, distinct and duplicate-value counts per column and complete formula warnings; no schema or business validity is inferred. Supply CSV OR columns/rows, never both. JSON numbers must be encoded as strings so identifiers keep leading zeros. One paid operation; no content retained. Reuse request_id only for identical input.
- `prepare_table`: Select, rename, derive and convert columns using a caller-owned version-1 recipe. Rules reference original input headers; output follows recipe order and omitted columns are listed. All input rows remain. Any failed conversion returns valid=false, all findings and csv_text=null, never a partially transformed file. No arithmetic, arbitrary code, schema inference or destination writes. One completed report costs one operation, including valid=false. Use reconcile_tables for two-file matching.
- `reconcile_tables`: Join two bounded tables using exact string or composite keys. Both sides must have non-empty unique keys: duplicates block the entire output, including many-to-many matches. Choose left, inner or full retention explicitly; unmatched records are reported even if excluded. Optional compare fields flag changed values without resolving them. Output includes row lineage; missing-side fields are empty strings. No trimming, fuzzy matching or destination writes. Combined inputs share 1 MB/10,000 records/250,000 cells. One completed report costs one operation; reuse request_id for identical retries. Full findings or an uncharged size error, never silent truncation.
- `validate_csv`: Check a CSV against your explicit schema; return all issue counts and at most 200 detailed findings. Headers match exactly; required=false permits empty cells, not missing headers. Unique compares raw non-empty strings. Dates use YYYY-MM-DD, decimals use a dot without grouping, booleans use true/false. A valid=false report is a completed paid operation, not a tool failure. No values are changed; use clean_csv for explicit transformations. Validity covers only the supplied rules, not destination-system or business correctness.
- `view_csv_demo`: Validate the fixed example with the current engine. Free and no arguments; demonstrates duplicate, email, amount and date findings. Custom files require validate_csv and paid access.
- `view_table_demo`: Join the bundled fixed tables and show changed and unmatched records. No arguments, key or credit. This cannot process custom data; use reconcile_tables with a paid key for your files.