Grok Ally
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Grok AllyGive me a second opinion on this design"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Grok Ally
Use Grok Build or Cursor Agent from Codex, Claude Code, or any local MCP client, with the selected service’s existing login. Keep context across messages, resume conversations, and cancel a turn when needed.
中文 · Usage reference · Changelog
Previously Grok Bridge. Upgrading from 0.2.x or earlier? Follow the migration steps.
Codex / Claude → MCP → Grok Ally → ACP → Grok Build / Cursor AgentInstall
You need Node.js 22+ and either Grok Build (grok login) or Cursor Agent (agent login). Only the selected backend needs to be installed. Use macOS or Linux; on Windows, run the client and bridge inside WSL.
Codex
codex plugin marketplace add yikZero/grok-ally
codex plugin add grok-ally@grok-allyStart a new Codex task after installation.
Claude Code
claude plugin marketplace add yikZero/grok-ally
claude plugin install grok-ally@grok-allyRestart Claude Code after installation. The same commands are available through /plugin in its interactive UI.
For other local clients, use the standard MCP configuration. All packages include the runtime; no npm install or build is needed.
Related MCP server: grok-oauth-mcp
Use it
Ask your agent: “Use Grok Ally to give me a second opinion on this design.” Then ask a follow-up in the same Grok conversation.
For code review, ask “Use Grok to review my uncommitted changes.” The grok-review skill keeps the review read-only and asks for evidence, file locations, and any verification limits.
Tool | Purpose |
| Start or continue a conversation |
| Follow progress, read complete answers, or find recent requests |
| Stop a turn |
| Check the selected backend installation |
To use Cursor, ask “Use Grok Ally through Cursor to implement this change.” The agent passes provider: "cursor"; its default model is grok-4.7[reasoning_effort=xhigh,fast=false]. Omit provider to keep using Grok Build, or set GROK_ALLY_PROVIDER=cursor in the MCP server environment to make Cursor the default. Existing sessions keep their original backend. There is no automatic quota fallback. See backend selection.
The agent supplies your project path and keeps the returned session ID for follow-ups. Share the context you want Grok to see in the prompt; host chat history is not imported automatically.
Long tasks return compact progress by default, then the answer when finished. Turn completed is the selected agent ending the turn, not confirmed tool outcomes or independent validation. Detailed tool history and full text remain available on demand. See results and cancellation and efficiency measurements.
Read-only operation is the default. Set write: true only for authorized edits. Grok uses its OS sandbox; Cursor uses verified Ask/Agent modes and native policies, which are not the same OS-level write restriction. Both backends retain their own tools, hooks, and network settings. See permissions and lifecycle for details.
Development
npm ci --ignore-scripts
npm run build
npm testEdit src/, rebuild, then reinstall. Both plugins share the same runtime; avoid editing generated bundles or installed caches.
Architecture · Validation · Release guide
Independent integration; not affiliated with xAI, Cursor, OpenAI, or Anthropic. Apache-2.0 · Attributions.
This server cannot be deployed
Maintenance
Related MCP Connectors
- QuallaaOAuthcom.quallaa
Talk to your public-facing AI from any MCP client — Claude, ChatGPT, Cursor, Cline, Windsurf.
Use AI models for chat, image, and video generation from Claude Code and other MCP hosts.
Generate contextual prompts and reusable agent skills, evaluate prompts with the 16-dimension Prompt Score, and manage saved work in PromptDrive. Twelve MCP tools also provide authorized access to private Memory for source-grounded answers. Connect over Streamable HTTP using OAuth 2.1 and PKCE. Generation consumes account quota and automatically saves successful results; Memory access follows account permissions and plan limits.
Build and manage AI-native customer support agents from Claude or any MCP client.
Related MCP Servers
- AlicenseBqualityDmaintenanceEnables interaction with the Grok AI through an MCP server, supporting chat completions, text completions, embeddings, and model operations with streaming capabilities.540 npm7MIT
- AlicenseBqualityDmaintenanceEnables local MCP clients to interact with xAI Grok through OAuth PKCE, supporting chat, X search, image/video generation, TTS, and transcription.94 npm2MIT
- FlicenseAqualityBmaintenanceEnables using the xAI Grok CLI as an MCP sub-agent for code review, asking questions, and continuing conversations within MCP hosts like Claude Code.4-
- FlicenseNot gradedqualityCmaintenanceAutomates Grok web interface to provide chat, web search, and conversation management via MCP, supporting multiple accounts and local/remote deployment.-