Skip to main content
Glama

Orin MCP

Use Orin's four model chains from Claude, Cursor, VS Code, and other MCP clients through hosted Streamable HTTP or local stdio.

Install

Mint a scoped token in Orin Chat → Account → MCP tokens, then install from GitHub:

npx -y github:januththedev/orin-mcp
ORIN_MCP_TOKEN=… orin-mcp

The package runs prepare/prepack, so dist/cli.js is built before execution. The npm dry-run package test verifies that the binary is included.

Hosted endpoint:

https://mcp.orinai.org/api/mcp
Authorization: Bearer <scoped-token>

Related MCP server: WarungCyber MCP Server

Tools and scopes

  • orin_chat, orin_generate — chat:generate

  • orin_list_models and orin://models — models:read

  • orin_usage and orin://usage — usage:read

Core enforces the same scopes on its direct /api/mcp endpoint. A token cannot bypass MCP scope checks by calling Orin Core directly.

Presets

  • orin-cheap

  • orin-balanced

  • orin-thinking

  • orin-coding

MCP requests are bounded to 100 messages, 12,000 characters per message, and 64 KiB aggregate request size.

Configuration

Variable

Required

Purpose

ORIN_MCP_TOKEN

stdio

Scoped Orin MCP credential

ORIN_API

No

https://orinai.org; loopback requires ORIN_ALLOW_LOCAL_CORE=1

ORIN_MCP_PER_MIN

No

Per-credential limit, default 60/min

ORIN_MCP_TIMEOUT_MS

No

Backend timeout, capped at 30 seconds

Diagnostics always use stderr so stdio stdout remains valid JSON-RPC.

Development

npm ci
npm run platform:build
npm run typecheck
npm test
npm run test:package

Tests are hermetic and use only local fakes. No live Orin, model, or external network call is made.

Security

  • Tokens and provider secrets are never logged.

  • Core verifies revocation and exact token type/scopes.

  • HTTP accepts only a bounded bearer token shape.

  • Provider API URLs are fixed to the trusted Orin Core origin.

  • Local stdio logs cannot corrupt the MCP protocol stream.

  • Unknown tool input is rejected before backend access.

License

MIT

Related MCP Connectors

Related MCP Servers

  • A
    license
    A
    quality
    B
    maintenance
    Enables MCP clients like Claude Code and Cursor to use multiple AI models (Gemini, GPT, Grok, DeepSeek, Kimi, Ollama) via a unified chat tool with conversation memory.
    3
    1
    Apache 2.0
  • A
    license
    A
    quality
    A
    maintenance
    Connects MCP clients such as Claude Desktop, Cursor and Claude Code to an OpenAI- and Anthropic-compatible AI gateway, enabling chat with Claude, GPT, Gemini, Grok and Kimi models, listing available models, and looking up per-million-token pricing. It also sends one prompt to several models in parallel and returns each answer with token usage, latency and estimated cost.
    4
    MIT
  • A
    license
    Not graded
    quality
    A
    maintenance
    Enables MCP-compatible clients to manage, run, and interact with local or cloud LLM models through exposed tools for listing, pulling, generating, chatting, embeddings, and web search/fetch. Supports Claude Desktop, Cline, and hybrid local/cloud setups.
    897 npm
    AGPL 3.0