Derive a voice profile from writing

Measures how someone already writes and returns a reusable voice profile: tone sliders, traits, and concrete rewrite rules with a before and after.

It reads writing that already exists rather than asking how someone would like to sound. The output is rules a writer or another model can apply, not adjectives to admire.

There is a page for people too: the brand voice tool runs the same engine in a browser, free and without an account.

Inputs and bounds

These bounds are the ones the server enforces. A request outside them is refused with a message naming the field.
Field	What it is	Required	Length
brand	Your brand or your name	required	2 to 80
samples	Two or three things you have written	required	40 to 4000
tone	How you want to sound	optional	up to 160

What comes back

sliders and traits - measured from the samples, not chosen from a list
rules - concrete, applicable instructions rather than descriptions
before and after - the same neutral paragraph rewritten by those rules, so the profile can be checked
metrics - the sentence and word statistics the reading is based on

Where the data goes

The samples are NOT sent anywhere. Only the derived rules and a fixed neutral paragraph reach the model provider, so unpublished writing stays on this server.

What it will not do

Not a plagiarism or authorship detector, and not a judgement of whether the writing is good. It describes the voice it is given.

Calling it

The server speaks MCP over JSON-RPC 2.0 at https://repic.site/api/free-tools/mcp. There is no key and no account. Calls share one per-IP daily budget with the free web tools, so an agent gets the same allowance a person does.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "derive_brand_voice",
    "arguments": {
      "brand": "Example Co",
      "samples": "We shipped the new editor today. It is faster, and it forgets nothing.\n\nYou asked for undo on every step. It is there now.",
      "tone": "Plain, direct"
    }
  }
}

Call tools/list first for every capability and its schema. The result carries both a readable content block and a lossless structuredContent object, so nothing has to be parsed out of prose.

The other skills

Brand persona

Builds the persona of a brand: the character it shows when it writes. Returns an archetype, voice traits, tone sliders, say and do-not-say lists, and a sample caption in that voice.

Brand kit

Produces a starter kit: a palette with contrast-checked pairings, a type pairing, voice, values and a mission line.

Brand consistency

Scores two or more recent posts for drift against each other and against a stated palette and voice, and returns per-check findings with specific fixes.

Social bio

Writes bio variants that fit the platform character limit, with the name field and the call to action handled separately.

Carousel outline

Turns a topic into a slide-by-slide carousel outline: a hook, one idea per middle slide, a specific ask, and review notes saying what is still weak.

Join early access →