EVOKORE
A unified Model Context Protocol (MCP) server that aggregates, indexes, and dynamically serves over 200+ specialized Agent Skills, Prompts, and Coding Workflows to AI assistants.
claude mcp add --transport stdio mattmre-evokore-mcp node /absolute/path/to/EVOKORE-MCP/dist/index.js \ --env ELEVENLABS_API_KEY="your_key_here" \ --env EVOKORE_TOOL_DISCOVERY_MODE="legacy | dynamic (default: legacy)" \ --env GITHUB_PERSONAL_ACCESS_TOKEN="your_token_here"
How to use
EVOKORE-MCP provides a single standard input/output (stdio) MCP endpoint that aggregates EVOKORE’s native workflow tools with proxied child MCP servers. It offers native tooling such as docs_architect, skill_creator, resolve_workflow, search_skills, get_skill_help, and discover_tools, while also routing to proxied servers defined in mcp.config.json (for example github, fs, and an optional elevenlabs server). Tools can be prefixed to avoid collisions (e.g., github_someTool) and discovery modes determine which tools are visible by default. In legacy mode you get the full native + proxied tool list; in dynamic mode native tools remain visible while proxied tools can be activated per session via discover_tools. When a protected proxied tool is invoked, EVOKORE returns an _evokore_approval_token and requires explicit human approval before retrying.
How to install
Prerequisites:
- Node.js and npm installed on your system
- Access to the EVOKORE-MCP repository and build artifacts
- Install dependencies and build
npm ci
npm run build
- Prepare environment
- Copy the example environment and customize values
cp .env.example .env
- Edit .env to include required credentials (or provide them via mcp config env vars):
GITHUB_PERSONAL_ACCESS_TOKEN=your_token_here
ELEVENLABS_API_KEY=your_key_here
# Optional behavior flag
EVOKORE_TOOL_DISCOVERY_MODE=legacy
# or
EVOKORE_TOOL_DISCOVERY_MODE=dynamic
- Register EVOKORE with your MCP client by providing the MCP server configuration
{
"mcpServers": {
"evokore-mcp": {
"command": "node",
"args": ["/absolute/path/to/EVOKORE-MCP/dist/index.js"]
}
}
}
-
Start the MCP router (as appropriate for your deployment). For local development, start the server and ensure it loads the environment variables from your .env or the mcpServers config.
-
Optional: use provided sync helpers if you expose CLI tooling
npm run sync:dry
npm run sync
Additional notes
Tips and notes:
- The mcp.config.json (or inline mcpServers config) defines proxied child servers such as github, fs, and an optional elevenlabs backend. Ensure credentials for these services are provided via environment variables or your deployment environment.
- In dynamic mode, you can reveal proxied tools per session using discover_tools. Exact-name compatibility remains for hidden proxied tools, meaning you can still invoke them by their exact prefixed name.
- HITL and approval workflows rely on _evokore_approval_token with one-time use and short retry windows. Ensure your operators are prepared to approve or reject actions accordingly.
- Windows runtime behavior: EVOKORE remaps only npx to npx.cmd; uv and uvx must resolve directly on PATH. If you use npx, be aware of platform-specific nuances.
- Ensure the environment provides the required tokens: GITHUB_PERSONAL_ACCESS_TOKEN for GitHub-backed proxied tools and ELEVENLABS_API_KEY for ElevenLabs capabilities if you enable that proxy.
Related MCP Servers
janee
Secrets management for AI agents via MCP • @janeesecure
mongo
MCP server that provide tools to LLMs such as claude in cursor to interact with MongoDB
vscode-context
MCP Server to Connect with VS Code IDE
attio
Model Context Protocol server for Attio CRM - enables Claude Desktop and Cursor to interact with your Attio workspace
mermaid-live
MCP server from iishyfishyy/mermaid-live-mcp
MCP-Swarm
Universal AI Agent Coordination Platform — 54 smart tools for multi-agent teamwork