MetaMCP Admin MCP
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., "@MetaMCP Admin MCP list all my MetaMCP instances"
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.
MetaMCP Admin MCP
An MCP server for administering MetaMCP instances — manage servers, namespaces, endpoints, and API keys across multiple MetaMCP deployments from any MCP client.
This is not the MetaMCP proxy (which aggregates tools). This is an admin tool that lets AI agents manage MetaMCP itself: add/remove servers, create namespaces, toggle tools, compare instances, and more.
Features
Multi-instance: Manage multiple MetaMCP deployments from a single MCP server
Full CRUD: Create, read, update, delete servers, namespaces, endpoints, and API keys
Bulk import: Import servers in Claude Desktop JSON format
Cross-instance ops: Compare servers across instances, clone server configs between instances
Health checks: Verify connectivity to all instances
Tool management: Inspect tools per server, toggle tools/servers within namespaces
Available Tools (22)
Tool | Description |
| List all configured MetaMCP instances |
| Check connectivity to all instances |
| Compare servers/namespaces/endpoints across instances |
| List all MCP servers on an instance |
| Get server details by UUID |
| Create a new MCP server |
| Update an existing server |
| Delete a server |
| Import servers from Claude Desktop JSON format |
| Copy a server config from one instance to another |
| List all namespaces |
| Get namespace details with servers |
| Create a new namespace |
| Update a namespace |
| Delete a namespace |
| List tools in a namespace |
| Enable/disable a server in a namespace |
| Enable/disable a specific tool in a namespace |
| List all endpoints |
| Get endpoint details |
| Create a new endpoint |
| Delete an endpoint |
| List all API keys |
| Create a new API key |
| Delete an API key |
| List tools for a specific server |
Installation
npm install -g metamcp-admin-mcpOr clone and build:
git clone https://github.com/danielrosehill/metamcp-admin-mcp.git
cd metamcp-admin-mcp
npm install
npm run buildConfiguration
Single instance
export METAMCP_BASE_URL="http://localhost:12008"
export METAMCP_INSTANCE_NAME="desktop" # optional
export METAMCP_SESSION_COOKIE="better-auth.session_token=..." # from browserMultiple instances
export METAMCP_INSTANCES='[
{
"name": "desktop",
"baseUrl": "http://localhost:12008",
"sessionCookie": "better-auth.session_token=..."
},
{
"name": "vm",
"baseUrl": "https://mcp.example.com",
"sessionCookie": "better-auth.session_token=..."
}
]'Getting the session cookie
MetaMCP uses Better Auth for session management. To get your session cookie:
Open MetaMCP in your browser
Open DevTools (F12) → Application → Cookies
Copy the
better-auth.session_tokenvalueUse it as
sessionCookiein config (prefix withbetter-auth.session_token=)
Claude Desktop / Claude Code config
{
"mcpServers": {
"metamcp-admin": {
"command": "node",
"args": ["/path/to/metamcp-admin-mcp/dist/index.js"],
"env": {
"METAMCP_INSTANCES": "[{\"name\":\"desktop\",\"baseUrl\":\"http://localhost:12008\",\"sessionCookie\":\"better-auth.session_token=...\"}]"
}
}
}
}How it works
MetaMCP exposes a tRPC API for its admin frontend. This MCP server acts as a tRPC client, calling those same endpoints with session cookie authentication. This means you get the same capabilities as the MetaMCP web UI, but accessible from any MCP client.
API Coverage
Built against the MetaMCP tRPC API:
frontend.mcpServers.*— Full CRUD + bulk importfrontend.namespaces.*— Full CRUD + tool/server status managementfrontend.endpoints.*— Full CRUDfrontend.apiKeys.*— Create, list, deletefrontend.tools.*— Query by server
License
MIT
Available Tools
26 toolsbulk_import_serversB
Bulk import MCP servers (Claude Desktop JSON format)
Inputs: servers (required), instance.
clone_serverC
Copy an MCP server config from one instance to another
Inputs: serverUuid (required), sourceInstance (required), targetInstance (required).
compare_instancesB
Compare MCP servers across all configured MetaMCP instances
Inputs: no named parameters.
create_api_keyC
Create a new API key
Inputs: name (required), instance.
create_endpointC
Create a new endpoint exposing a namespace
Inputs: name (required), namespaceUuid (required), description, enableApiKeyAuth, +2 more.
create_namespaceC
Create a new namespace
Inputs: name (required), description, mcpServerUuids, instance.
create_serverC
Create a new MCP server on a MetaMCP instance
Inputs: name (required), type (required), description, command, +6 more.
delete_api_keyC
Delete an API key
Inputs: uuid (required), instance.
delete_endpointD
Delete an endpoint
Inputs: uuid (required), instance.
delete_namespaceC
Delete a namespace
Inputs: uuid (required), instance.
delete_serverB
Delete an MCP server from a MetaMCP instance
Inputs: uuid (required), instance.
get_endpointC
Get endpoint details
Inputs: uuid (required), instance.
get_namespaceB
Get namespace details including its servers
Inputs: uuid (required), instance.
get_namespace_toolsC
List tools available in a namespace
Inputs: namespaceUuid (required), instance.
get_serverB
Get details of a specific MCP server by UUID
Inputs: uuid (required), instance.
get_server_toolsB
List tools registered for a specific MCP server
Inputs: mcpServerUuid (required), instance.
health_checkA
Check connectivity to all configured MetaMCP instances
Inputs: no named parameters.
list_api_keysB
List all API keys on a MetaMCP instance
Inputs: instance.
list_endpointsC
List all endpoints on a MetaMCP instance
Inputs: instance.
list_instancesC
List all configured MetaMCP instances
Inputs: no named parameters.
list_namespacesC
List all namespaces on a MetaMCP instance
Inputs: instance.
list_serversA
List all MCP servers on a MetaMCP instance
Inputs: instance.
toggle_server_in_namespaceB
Enable or disable a server within a namespace
Inputs: namespaceUuid (required), serverUuid (required), status (required), instance.
toggle_tool_in_namespaceC
Enable or disable a specific tool within a namespace
Inputs: namespaceUuid (required), toolUuid (required), serverUuid (required), status (required), +1 more.
update_namespaceC
Update an existing namespace
Inputs: uuid (required), name (required), description, mcpServerUuids, +1 more.
update_serverC
Update an existing MCP server
Inputs: uuid (required), name (required), type (required), description, +7 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 26 tools
Tools are clearly partitioned by resource type (servers, namespaces, endpoints, api_keys, instances) with distinct action verbs preventing overlap. Minor density exists between get_namespace and list_namespaces, but descriptions clarify that one retrieves detailed server membership while the other provides a simple listing.
Excellent adherence to snake_case with a strict verb_noun pattern throughout all 26 tools (create_server, delete_namespace, list_api_keys, etc.). Compound verbs like bulk_import_servers and health_check follow the same convention without mixing styles or cases.
With 26 tools, the surface exceeds the typical ideal range (3-15) and enters the heavy category (16-25+). While each tool earns its place covering full CRUD for four distinct resource types plus specialized operations (clone, toggle, import), the breadth creates a dense interface that approaches cognitive overload.
Provides comprehensive lifecycle coverage for servers and namespaces, including granular controls like toggling specific tools. Minor gaps exist for endpoints (missing update_endpoint) and API keys (lacking get single key), though these can be worked around via list operations.
Related MCP Connectors
MCP server for your apps' tools and custom tools, plus hosted AI agents and approval-gated workflows
Your org's AI agents, tasks, runs, search, and brain files as MCP tools and resources.
214 tools for AI agents over one MCP endpoint: image, video, audio, 3D, data, storage, mail, GPUs.
Secure and manage MCP servers through the Speakeasy AI Control Plane from your agent.