Skip to main content
A Python framework that extends standard MCP servers with production capabilities. Drop-in replacement for FastMCP:change one import and you get:

​
Quick Reference

from concierge import Concierge, Config, ProviderType

# Create a server
app = Concierge("my-server")

# Register tools
@app.tool()
def my_tool(param: str) -> dict:
    return {"result": "done"}

# Define workflow stages
app.stages = {
    "step1": ["my_tool"],
    "step2": ["other_tool"],
}

# Define allowed transitions
app.transitions = {
    "step1": ["step2"],
    "step2": [],
}

# Use state across stages
app.set_state("key", "value")
value = app.get_state("key", "default")

​
Sections

Transitions

Stages, transitions, and progressive tool disclosure

State

Per-session persistent state with snapshotting

Backends

Plain, Search, Plan, and Code execution modes

Tools

Define and expose tools to AI agents

Resources

Expose read-only data to clients

Widgets

Interactive UI rendered in AI conversations