Grammar Check Skill

Overview

Grammar Check sends your `text` to a helper service built on harper-core 3, the Rust library behind the open-source Harper grammar checker, with an APIVerve layer on top that tunes it for automatic correction. Leading and trailing whitespace is trimmed first; line breaks inside the text are kept. An empty string returns a 400 error, and an input the checker service rejects also returns a 400 rather than a 500.

The text is parsed as Markdown, so inline code, links and URLs are not checked. The English variant is detected from the spelling the writer already uses and returned in `dialect`; British, Canadian, Australian and Indian spellings such as colour or theatre are not changed to American ones. Harper's grammar rules then run over the text, covering subject-verb agreement, a/an, their/they're/there, its/it's, your/you're, could of, repeated words, sentence capitalisation and several hundred common word-choice mistakes. Wording and style advice (shorter phrasings, en dashes and similar preferences) is not reported, because applying it automatically would rewrite correct text.

Spelling suggestions are ranked by how people actually mistype: swapped letters, a missing or extra double letter, a wrong vowel, a neighbouring key and words that sound alike count as smaller mistakes than arbitrary edits, and common words rank ahead of rare ones. A word typed in lowercase is never corrected to a proper noun, and the case you typed is kept, so TOMOROW becomes TOMORROW. Capitalised words in the middle of a sentence are treated as names and left alone unless they are one slip away from a known place or name. Acronyms, words with digits, camelCase identifiers and common technical terms such as JSON, OAuth or npm are not flagged.

`errorCount` is the number of reported problems and `modified` is true when there is at least one. `corrected` is built by applying the first suggestion of every problem, working from the end of the text backwards so earlier offsets stay valid; overlapping problems are reduced to one first, so every correction applies cleanly.

Each entry in `corrections` carries `rule` (the kind of problem, for example `Spelling`, `Agreement`, `Repetition`, `Capitalization` or `Word Choice`), `message`, `position.start` and `position.end` as character offsets into the trimmed text, and `suggestions`, the replacement strings for that span, best first. An empty-string suggestion means delete the span. Offsets count characters, not bytes, so they line up with JavaScript string indexes for most text. On the Free plan `corrections` is null and only the summary fields are returned.

Live Test Grammar Check Skill Skill →

The tool

Once your client is connected to the VerveKit server, this appears in its tool list as GrammarCheckSkill. It is read-only and open-world — it fetches and never mutates anything on your side — so most clients call it without asking you to confirm.

{
  "name": "GrammarCheckSkill",
  "arguments": {
    "text": "This is a sample text with a grammatical error."
  }
}

You do not name the tool yourself; the model picks it. Asking about This is a sample text with a grammatical error. in the terms this skill covers is enough for it to reach for GrammarCheckSkill on its own — naming it explicitly also works, and is the way to force the call.

Connecting

One server URL covers every skill in the catalog, including this one. Authorization is OAuth: the client opens a browser once, and there is no key to paste into a config file.

{
  "mcpServers": {
    "vervekit": {
      "url": "https://api.vervekit.com/v1/mcp"
    }
  }
}

https://api.vervekit.com/v1/mcp

Per-client setup — Claude, Cursor, VS Code, ChatGPT — is on the MCP setup page.

Arguments

These are the properties on the tool's inputSchema, so a well-behaved client validates them before the call is made. Premium arguments are accepted on every plan but only take effect on plans that include them.

What the model gets back

The result carries a structuredContent object matching the tool's declared outputSchema, so a client reads fields without parsing prose. status is "ok" and error is null on success; a null field means the value was not available for that input, not that the call failed.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46

{
  "status": "ok",
  "error": null,
  "data": {
    "modified": true,
    "corrected": "The pig snorted when I saw it on the barn gate. I was a little mad because it woke up my puppy",
    "errorCount": 3,
    "dialect": "American",
    "corrections": [
      {
        "rule": "Repetition",
        "message": "Did you mean to repeat this word?",
        "position": {
          "start": 0,
          "end": 7
        },
        "suggestions": [
          "The"
        ]
      },
      {
        "rule": "Spelling",
        "message": "Did you mean `because`?",
        "position": {
          "start": 71,
          "end": 78
        },
        "suggestions": [
          "because"
        ]
      },
      {
        "rule": "Spelling",
        "message": "Did you mean `puppy`?",
        "position": {
          "start": 93,
          "end": 101
        },
        "suggestions": [
          "puppy",
          "puppetry"
        ]
      }
    ]
  }
}

Response fields

Paths are relative to data. Premium fields are absent rather than zeroed on plans that do not include them, so check for presence instead of comparing to 0.

Failure modes

Errors come back as tool errors carrying a sentence the model can act on, not a bare status code. Error handling covers the full list.

A call costs 20 credits each time the tool actually runs; a model that reasons about the tool without calling it costs nothing.

Use cases

Content Management Proofing

Flag grammar mistakes in submitted blog drafts and substitute corrected sentences before publishing articles to readers.

Customer Support Messaging

When agents write replies in ticketing desks, the system uses the error count to warn them of uncorrected typos.

Resume Intake Screening

Job portals scan cover letters during applicant uploads, noting whether the text was modified to preserve clean candidate profiles.

Student Essay Feedback

Learning management platforms pass student submissions to display the number of errors and return properly formatted text.

Other ways to use Grammar Check Skill

Set up Grammar Check Skill on VerveKit, or reach the same source a different way. Your VerveKit account and credits work on all of them — one key, one balance.

Related

More in Text Processing: