feat(cli): markkit mcp stdio server exposing the markkit tool

CHANGELOG.md

@@ -9,6 +9,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/).
 
 ### Added
 
+- `markkit mcp`: Model Context Protocol server over stdio exposing the `markkit` tool
+  (`tools/list`, `tools/call`). URL fetching keeps the SSRF guard; `--no-url` keeps it offline,
+  the `extract` limit flags and `--max-output-chars` bound each call.
 - Plain text input (`text/plain`, `.txt`), always on: paragraphs and hard breaks through the IR,
   UTF-8/UTF-16 BOM decoding, control characters dropped, `invalid_utf8` warning on lossy decode.
 - Markdown input (`markdown` feature, in `all_formats`; `text/markdown`, `.md`): re-parsed with

README.md

@@ -91,8 +91,13 @@ markkit extract --file page.html --base-url https://example.com/blog/
 
 # From stdin
 cat page.html | markkit extract -
+
+# MCP server (stdio) exposing the `markkit` tool to agents
+markkit mcp            # add --no-url to keep it offline
 ```
 
+MCP client config: `{"mcpServers": {"markkit": {"command": "markkit", "args": ["mcp"]}}}`.
+
 ### Python
 
 ```python

crates/markkit-cli/Cargo.toml

@@ -27,8 +27,8 @@ spreadsheet = ["markkit/spreadsheet"]
 markdown = ["markkit/markdown"]
 
 [dependencies]
-markkit = { path = "../markkit", version = "0.1.1", features = ["http_client"] }
-tokio = { workspace = true, features = ["macros", "net", "rt", "rt-multi-thread", "time", "fs"] }
+markkit = { path = "../markkit", version = "0.1.1", features = ["http_client", "tool"] }
+tokio = { workspace = true, features = ["macros", "net", "rt", "rt-multi-thread", "time", "fs", "io-std", "io-util"] }
 clap.workspace = true
 anyhow.workspace = true
 serde_json.workspace = true

crates/markkit-cli/src/main.rs

@@ -6,6 +6,8 @@ use std::time::Duration;
 use clap::{Args, Parser, Subcommand};
 use markkit::{Limits, Markkit};
 
+mod mcp;
+
 #[derive(Parser)]
 #[command(
     name = "markkit",
@@ -53,6 +55,27 @@ enum Commands {
         #[arg(long)]
         main_content: bool,
 
+        #[command(flatten)]
+        limits: LimitArgs,
+    },
+    /// Serve the `markkit` tool over MCP (JSON-RPC on stdin/stdout)
+    Mcp {
+        /// Do not offer the `url` source (no network access)
+        #[arg(long)]
+        no_url: bool,
+
+        /// Maximum markdown characters returned per call
+        #[arg(long, default_value_t = markkit::tool::DEFAULT_MAX_OUTPUT_CHARS)]
+        max_output_chars: usize,
+
+        /// Keep only the detected main content (article body) of HTML pages
+        #[arg(long)]
+        main_content: bool,
+
+        /// Locale for tool descriptions and errors (en-US, uk-UA)
+        #[arg(long, default_value = "en-US")]
+        locale: String,
+
         #[command(flatten)]
         limits: LimitArgs,
     },
@@ -133,6 +156,24 @@ async fn main() -> Result<()> {
     let cli = Cli::parse();
 
     match cli.command {
+        Commands::Mcp {
+            no_url,
+            max_output_chars,
+            main_content,
+            locale,
+            limits,
+        } => {
+            let limits = limits.apply(Limits::default());
+            let max_message = mcp::max_message_bytes(limits.max_input_size);
+            let tool = markkit::tool::MarkkitTool::builder()
+                .locale(&locale)
+                .config(Markkit::builder().limits(limits).main_content(main_content))
+                .allow_url(!no_url)
+                .max_output_chars(max_output_chars)
+                .build();
+            let stdin = tokio::io::BufReader::new(tokio::io::stdin());
+            mcp::serve(&tool, stdin, tokio::io::stdout(), max_message).await?;
+        }
         Commands::Extract {
             source,
             file,
@@ -215,7 +256,9 @@ mod tests {
 
     fn limits_from(args: &[&str]) -> Limits {
         let cli = Cli::try_parse_from(["markkit", "extract", "x"].iter().chain(args)).unwrap();
-        let Commands::Extract { limits, .. } = cli.command;
+        let Commands::Extract { limits, .. } = cli.command else {
+            panic!("expected extract");
+        };
         limits.apply(Limits::default())
     }