criticmarkup
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.
You are an editorial review assistant using the CriticMarkup MCP server.
CRITICAL PROTOCOL RULES:
1. IMMUTABLE DIFFS: CriticMarkup tags ({~~...~~}, {--...--}, {++...++}) are read-only markers
reserved strictly for the human author to accept or reject in their editor (e.g., Monoleaf).
Preserve human governance and accountability.
2. NEVER REVERT OR CLEAN DIFFS: Never include CriticMarkup delimiters in 'search_string' or 'replacement_string'.
Do not attempt to undo, revert, clean up, or accept previous edits.
3. SURGICAL REVISIONS: Keep search and replacement strings minimal (under 15 words). Do not replace whole paragraphs.
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
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| get_md_contentsA | Reads a markdown file, returning prose up to the user-configured cutoff line (if set), followed by an abbreviated summary of discussion comment threads. YAML frontmatter is completely hidden from view. |
| replace_multipleB | Applies multiple targeted revisions across different sections in a single call using CriticMarkup. Edits are applied bottom-to-top to maintain offset stability. Targeted edits past the configured cutoff line are rejected. |
| request_changesB | Applies a single targeted change or deletion using CriticMarkup. Targeted edits past the configured cutoff line are prohibited. |
| replace_allB | Globally replaces ALL instances of a word or phrase throughout the document body using CriticMarkup. Operates across the full document body regardless of cutoff line, while frontmatter, inline anchors, and JSON comment payloads remain strictly protected. |
| write_commentB | Attaches an inline comment anchor around target text and creates an open discussion thread. Comments cannot be anchored past the configured cutoff line. |
| reply_to_commentA | Replies to an existing discussion thread by its 4-character UID (e.g. 'o1b7'). Unaffected by the line cutoff. |
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 6 tools
Most tools have clearly distinct purposes: reading, targeted single/multiple edits, global replace, creating comments, and replying. However, request_changes applies a single edit despite a plural name that suggests multiple changes, and it overlaps somewhat with replace_multiple at different granularities, so an agent could momentarily confuse them.
All tool names use snake_case with a verb_noun structure (get_md_contents, replace_multiple, request_changes, replace_all, write_comment, reply_to_comment). The naming pattern is predictable throughout; the semantic oddity of request_changes does not break the naming convention.
Six tools is well-scoped for markdown reading, editing, and commenting. Each tool appears to earn its place without excessive granularity or redundancy in count.
Core read/edit/comment-creation workflows are covered, including single, multiple, and global edits plus replying to threads. But the discussion lifecycle is incomplete: there is no resolve, close, or delete comment-thread tool, and no accept/reject lifecycle for CriticMarkup changes, leaving agents at a dead end for closing or cleaning up discussions.