MCP Server Overview

Floci Studio ships with a built-in Model Context Protocol (MCP) server. It exposes 112 tools across 19 service modules (plus any community plugins), giving any MCP-compatible AI client full operational control over your local AWS environment.

What you can do

Once connected, your agent can:

• Create infrastructure — queues, topics, buckets, tables, functions, secrets
• Operate services — send SQS messages, invoke lambdas, publish SNS events, put S3 objects
• Read state — list resources, receive messages, scan DynamoDB, get CloudWatch logs
• Orchestrate flows — start Step Functions executions, put EventBridge events, run Athena queries
• Manage secrets — create/read/update Secrets Manager and KMS entries
• Marketplace — deploy and teardown recipes from conversation (including DynamoDB Admin)
• Tag resources — apply, remove, and search tags across all AWS resource types
• Generate artifacts — export Terraform, generate architecture Mermaid diagrams, create JWT tokens

Architecture

AI Agent (Claude / Cursor)
        │  JSON-RPC stdio
        ▼
floci MCP Server (Python, FastMCP)
        │  in-memory httpx.ASGITransport
        ├──► FastAPI sidecar (Lambda, Athena, Marketplace)
        │
        │  boto3 direct
        └──► floci engine :4566 (SQS, SNS, S3, DynamoDB, ...)

The MCP server is built with FastMCP. Each service module exports a register(mcp) function that decorates async functions as tools using @mcp.tool().

Tool modules

Module	Tools	Description
meta	6	Health checks, service inventory, architecture diagram
lambda_	8	Create, invoke, update, delete functions
sqs	8	Queue CRUD, send/receive/delete/purge messages
sns	8	Topic CRUD, publish, subscribe/unsubscribe
s3	8	Bucket CRUD, object get/put/delete, presigned URLs
dynamodb	8	Table CRUD, put/get/query/scan items
secrets	5	Secrets Manager CRUD
kms	5	Key management, encrypt/decrypt
eventbridge	5	Bus list, put events, rule CRUD
stepfunctions	4	List, start, describe executions
athena	3	Run queries, list databases, query history
ses	4	Verify identities, send emails, quota
marketplace	5	List, deploy, teardown recipes (incl. DynamoDB Admin)
devtools	6	Terraform export, AWS CLI escape hatch, JWT, proxy
observability	8	DLQ redrive, flight recorder (time-travel), service graph
iac	2	IaC auto-discovery and drift detection
hybrid	5	Cloud seeding, live SQS proxy, reverse tunnels
extensibility	8	Lifecycle webhooks, HTTP interceptors, plugin catalog
tags	6	Tag/untag resources, search by tag, list all tag keys

Example agent conversation

You: Create an SQS FIFO queue named “order-events”, send three test orders to it, then create a Lambda function in Python that reads from the queue and logs the order IDs.

The agent will:

1. Call create_sqs_queue(name="order-events", fifo=True)
2. Call send_sqs_message(...) three times
3. Call create_lambda_function(name="order-processor", runtime="python3.12", ...)
4. Call create_event_source_mapping to wire the queue to the function
5. Call get_lambda_logs to show you the output

All in a single conversation turn.

Next steps

• Connect Claude or Cursor
• Full tools reference