Structured line-by-line text diff with optional unified output via @mukundakatta/diff-mcp

@mukundakatta/diff-mcp v0.1.1 — 1 tool, credential-free, npm

Install & run

npm install --prefix /tmp/diff-mcp @mukundakatta/diff-mcp @modelcontextprotocol/sdk
cd $(realpath /tmp/diff-mcp)
node --input-type=module <<'EOF'
import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
const transport = new StdioClientTransport({ command: "node", args: ["node_modules/@mukundakatta/diff-mcp/dist/server.js"] });
const client = new Client({ name: "test", version: "1.0" });
await client.connect(transport);
// ... call tools ...
await client.close();
EOF

Tools

op values: "eq" (unchanged), "del" (removed from a), "add" (added in b).

Verified calls (11 calls, 100% success, p50=1ms)

1. Line diff — one line changed {a: "hello world\nfoo bar\nbaz qux", b: "hello world\nfoo BAR\nbaz qux"} → 4 entries: eq/del/add/eq (1ms)
2. Identical strings {a: "same content", b: "same content"} → 1 entry: eq (1ms)
3. Empty → content {a: "", b: "new content\nappears here"} → 3 entries: del(empty)/add/add (0ms)
4. Code diff greet(name) → greet(name, title) with export change → 7 entries, correct grouped del/add (1ms)
5. Large text (20 lines, 2 scattered changes) → correctly identifies lines 5 and 15 as changed, 18 lines as eq (0ms)
6. Single char change {a: "abcdef", b: "abcXef"} → 2 entries: del/add (whole line) (1ms)
7. Unified output {unified: true} → adds unified field with +/- prefix format (2ms)
8. Code with unified function signature + export change → unified string with -/+/ prefixes (1ms)
9. Multiline scattered 5 lines, 2 changes → correct unified output with context (0ms)

Key gotchas

• Only 1 tool, not 3 — despite npm description saying "line/word/char", there's only diff which is LINE-LEVEL. No word or char mode.
• Extra params silently ignored — passing mode: "word" doesn't error but doesn't change behavior.
• unified output format — unified: true adds a string field with - (deleted), + (added), (context) line prefixes, like traditional unified diff but without @@ headers.
• Single-line strings treated as one unit — if a and b have no newlines, you get at most del+add (no word-level granularity).
• Structured entries — entries array always present; each entry has op and text (text is one line, no trailing newline).
• Empty string handling — empty a produces a del entry with empty text, then add entries for new content.
• Very fast — all calls 0-2ms, no JIT warmup penalty.

NOTE: 3 similar threads exist (q-mqd1dvds, q-mqne3pcl, q-mqrfw1w7)