home / mcp / grammarly mcp server
A single-tool MCP server that analyzes Grammarly AI detection and plagiarism scores via browser automation and text rewriting.
Configuration
View docs{
"mcpServers": {
"bjornmelin-grammarly-mcp": {
"command": "node",
"args": [
"/path/to/grammarly-mcp/dist/server.js"
],
"env": {
"BROWSER_PROVIDER": "stagehand",
"BROWSERBASE_API_KEY": "bb_xxx",
"BROWSERBASE_CONTEXT_ID": "ctx_xxx",
"BROWSERBASE_PROJECT_ID": "xxx"
}
}
}
}You can run a self-contained MCP server named Grammarly that automates Grammarlyβs web UI to assess AI-detection and plagiarism scores, and then rewrites text to reduce detection using Claude. You access it through a simple MCP client workflow and control it via environment variables. This enables automated scoring, optional rewriting, and structured outputs for your text optimization needs.
Use an MCP client to connect to the Grammarly MCP Server and invoke the grammarly_optimize_text tool. Set your environment to use Stagehand with Browserbase by default for reliable sessions and persistent login contexts, or switch to Browser Use Cloud if you already have that setup. Provide the text you want analyzed and, if needed, specify the rewriting preferences such as tone, domain, and iterations. The server will first score the original text, then, if you enable optimize mode, iterate rewriting and re-scoring until the thresholds for AI detection and plagiarism are met or until the maximum iterations are reached. The output includes the final text, scores, and a live URL for debugging when available.
Follow these concrete steps to install and run the Grammarly MCP Server locally.
Prerequisites you need on your system:
# 1. Clone the project
git clone https://github.com/BjornMelin/grammarly-mcp.git
cd grammarly-mcp
# 2. Install dependencies
pnpm install
# 3. Build the server
pnpm build
# 4. Run the server locally (stdio MCP transport)
pnpm start
# or
node dist/server.jsChoose a browser automation provider. The default is Stagehand + Browserbase. You can fall back to Browser Use Cloud if you have that setup. You must supply credentials for the chosen provider in your environment. The minimal, recommended setup uses Stagehand with Browserbase and includes your Browserbase API key, project ID, and the provider setting.
# Minimal example config (Stagehand + Browserbase)
{
"env": {
"BROWSER_PROVIDER": "stagehand",
"BROWSERBASE_API_KEY": "bb_xxx",
"BROWSERBASE_PROJECT_ID": "xxx"
}
}
```
```
# Optional: add context for persistent login (if you have this handy)
{
"env": {
"BROWSER_PROVIDER": "stagehand",
"BROWSERBASE_API_KEY": "bb_xxx",
"BROWSERBASE_PROJECT_ID": "xxx",
"BROWSERBASE_CONTEXT_ID": "ctx_xxx"
}
}Configure your MCP client to point at the local server. You will typically run the server with Node and pass the server script path as an argument. The client configuration can be done through interactive setup or by manual config in your MCP client of choice.
# Claude Code CLI example (manual)
claude mcp add grammarly -e BROWSER_PROVIDER=stagehand \
-e BROWSERBASE_API_KEY=bb_xxx \
-e BROWSERBASE_PROJECT_ID=xxx \
-- node /path/to/grammarly-mcp/dist/server.js
```
```
# Claude Desktop / other clients
# Add a server with the same env values and the same node command
{
"mcpServers": {
"grammarly": {
"command": "node",
"args": ["/path/to/grammarly-mcp/dist/server.js"],
"env": {
"BROWSER_PROVIDER": "stagehand",
"BROWSERBASE_API_KEY": "bb_xxx",
"BROWSERBASE_PROJECT_ID": "xxx"
}
}
}
}This server exposes a single tool named grammarly_optimize_text that analyzes and optimizes text for Grammarly AI detection and plagiarism scores. You can configure the mode, thresholds, tone, and iterations to control how aggressively rewriting occurs.
The server supports two browser automation providers. By default you use Stagehand with Browserbase. You can switch to Browser Use Cloud if you have that setup.
Environment variables you will commonly set when using Stagehand with Browserbase include the API key and project ID, along with the provider selection. You can also enable a persistent login context if desired.
Treat API keys and session data as sensitive. Do not commit credentials to version control. Use per-project or per-environment configurations, and consider context IDs for Browserbase to keep Grammarly login state across runs. Review data flow: automation server -> Grammarly web UI -> Claude for rewriting.
If the server fails to start, verify that required environment variables are set for your chosen provider. If a tool does not appear in your client, ensure the server path is correct and that you rebuilt the project after changes.
If you encounter login persistence issues with Browserbase, ensure you logged in while the context was active and that the context ID matches the session. For Browser Use Cloud, verify your API key and that the profile is properly synced.
UI automation depends on Grammarlyβs web UI stability. Stagehand provides self-healing capabilities, while Browser Use Cloud offers a different approach with per-task debugging. Respect usage limits of each provider and plan for potential adjustments when UI changes occur.
Tool that analyzes AI detection and plagiarism scores for given text and can rewrite text to reduce detection, with configurable mode, thresholds, tone, and iterations.