MCP Compact
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., "@MCP Compact invoke filesystem to list /tmp"
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.
MCP Compact
MCP Compact aggregates multiple upstream MCP servers into one stdio-only runtime with a single compact surface.
What It Is
MCP Compact runs as one MCP server over stdio and exposes only invoke and read.
Related MCP server: gtwmcp
Use From An MCP Client
MCP Compact runs as a stdio MCP server. It does not expose HTTP or /mcp.
Configure your MCP client to launch it like this:
{
"mcpServers": {
"mcp_compact": {
"command": "uv",
"args": ["run", "mcp-compact", "config.example.json"]
}
}
}If you want to start it manually, run:
uv run mcp-compact config.example.jsonUpstream Config
config.example.json is the config file consumed by mcp-compact itself:
{
"mcpServers": {
"filesystem": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"]
}
}
}All upstreams must use stdio.
Test
uv run ruff check src/mcp_compact tests
uv run mypy src/mcp_compact
uv run pytest tests/ -qAvailable Tools
2 toolsinvokeC
Invoke an MCP tool through the compact projection surface.
Use ref.server and ref.name with canonical upstream identifiers.
Use mode="validate" to inspect input requirements before calling.
Available tool families:
Example: invoke(ref={"server": "notes", "name": "echo_note"}, arguments={"title": "...", "body": "..."})
Inputs: ref (required), mode, arguments. Returns: result
readA
Read MCP resources through the compact projection surface.
Use ref.server with either ref.uri for direct resources or
ref.uriTemplate plus template arguments for resource templates.
Use mode="preview" before reading when you need metadata first.
Direct resources:
none
Resource templates:
none
Example: read(ref={"server": "notes", "uriTemplate": "memo://notes/{slug}", "arguments": {"slug": "welcome"}}, mode="read")
Inputs: ref (required), mode. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 2 tools
The two tools are clearly separated by purpose: invoke handles tool calls and read handles resource access. The descriptions provide explicit examples and modes to guide usage, though the generic proxy nature could occasionally blur the line between a tool and a resource.
Both tool names are single-word verbs in a consistent imperative style, making them predictable and easy to remember. There is no mixing of conventions or misleading synonyms.
With only two tools, the set is intentionally minimal but well-suited for a compact projection surface that forwards to a larger MCP ecosystem. While it feels thin compared to direct-integration servers, its purpose justifies the small count.
The server covers the core interaction patterns (invoking tools and reading resources) and includes validation/preview modes. However, it lacks an explicit discovery mechanism to list available servers, tools, or resource templates, which could be a significant gap for agents without prior knowledge.
Maintenance
Related MCP Connectors
Stateless MCP gateway and OTel span-streaming bridge for hosted MCP servers.
MCP Gateway: wrap any MCP server with cold-start retries, uptime SLA, and per-execution MPP billing.
Host any HTTP API as a standard MCP server; auto-distribute to 12+ mainstream agent platforms.
MCP server for Superserve sandboxes: create, exec, and manage Firecracker microVMs
Related MCP Servers
AlicenseNot gradedqualityBmaintenanceConsolidates multiple upstream MCP servers behind a single STDIO interface, exposing search_tools and run_tool to avoid context bloat.380 npm20MIT- AlicenseNot gradedqualityDmaintenanceAggregates multiple MCP servers behind a single stdio interface, supporting stdio-based and HTTP SSE-based upstreams with OAuth.4 npmMIT
- AlicenseNot gradedqualityDmaintenanceAggregates multiple upstream MCP servers (stdio, streamable-http, sse, websocket) behind a single streamable-HTTP endpoint with declarative configuration, namespacing, filtering, and auto-reload.MIT
- AlicenseNot gradedqualityCmaintenanceStarts local MCP backend servers only when their tools are called, exposes their tools through a single stdio gateway, preserves stateful sessions, and can stop safely restartable backends after idle timeouts.1MIT