Humanizer MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Use humanize_copy for text the user will publish in a website/app, include in an audience-facing document, or send to someone else. Ordinary assistant conversation stays unchanged. Review returned rewrites for meaning and factual accuracy before use. If disabled, use the original draft without retries. Only the user controls the menu bar switch.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| humanizer_statusA | Check whether local Humanizer is ready, loading, busy, off, or unavailable. |
| humanize_copyA | Rewrite audience-facing copy locally. Submit named text fields, with exact preserve terms and optional max_characters. Keep each call around 300 words total to fit the local time budget. Use plain for app strings/headlines, markdown for document prose. Markdown headings, tables, quotes, and code remain intact; extract headings as plain fields if they need rewriting. Returns originals on validation failure or when the user switches it off. Mechanical checks do not establish factual fidelity; review every accepted rewrite before use. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 2 tools
The two tools split cleanly along read-only status vs. mutating rewrite, with no overlap in intent or inputs. An agent can pick correctly from the names alone, and the descriptions reinforce the boundary (check readiness first, then submit fields).
Both names are snake_case and readable, but they use slightly different stems: 'humanizer_status' is a noun-prefixed resource, while 'humanize_copy' is a verb_noun action. It's a minor deviation in an otherwise predictable pair, not a real source of confusion.
Two tools is borderline thin, though it matches a narrow single-purpose local service (state check plus one operation). It earns its place, but with only a status probe and one action there is little to compose, so it sits at the low end of well-scoped.
The core lifecycle (is it ready → rewrite) is covered, but there is no way to cancel a busy run, batch/chunk long inputs despite the ~300-word guidance, or adjust tone/style beyond a plain-vs-markdown format flag. Agents can work around these manually, but the surface has notable gaps for anything beyond a single short call.