Technology

txn2/mcp-s3

An open-source Go MCP server that exposes S3 (and any S3-compatible store) as governed tools to AI agents — browse buckets, read/write objects, mint presigned URLs — designed as a composable library, not just a standalone binary.

5 connections 1 resources 1 post

Summary

What it is

An open-source Go MCP server that exposes S3 (and any S3-compatible store) as governed tools to AI agents — browse buckets, read/write objects, mint presigned URLs — designed as a composable library, not just a standalone binary.

Where it fits

The reference implementation of the Agentic Data Plane on object storage. It sits between an agent runtime and S3, translating natural-language intent into authenticated S3 operations with secure defaults.

Misconceptions / Traps
  • It is not a database connector — it hands the model raw object access, so the guardrails (read-only mode, GET size caps, prefix ACLs) are what keep an agent from overrunning the context window or writing where it shouldn't.
  • "MCP server" here means a tool surface, not a hosting service; you embed it in your gateway.
Key Connections
  • txn2/mcp-s3 extends Model Context Protocol (MCP) — implements the MCP tool contract for S3
  • txn2/mcp-s3 integrates_with MinIO, SeaweedFS — works against any S3-compatible backend
  • Composable counterpart to AIStor MCP Server and S3 Tables MCP Server

Definition

What it is

An open-source Model Context Protocol (MCP) server, written in Go, that lets AI assistants browse buckets, read and write objects, and generate pre-signed URLs against Amazon S3 and any S3-compatible store. It is designed as a composable library platform engineers can embed in custom AI gateways, not just a standalone server.

Why it exists

Multi-agent systems need just-in-time context loading from object storage, but raw S3 access via scripts or signed HTTP requests is brittle for an LLM to drive. mcp-s3 exposes S3 as governed MCP tools with secure defaults, turning object storage into an agent-accessible memory substrate.

Primary use cases

Agentic retrieval of JSON/Parquet/unstructured data from S3, building unified data-access layers into AI gateways, secure multi-tenant agent access to buckets.

Recent developments

Latest signals
  • Secure-by-default agent guardrails. Ships read-only modes, configurable GET size caps (e.g. 10 MB) to prevent context-window overflow, and prefix-based ACLs scoping agents to specific paths. Per GitHub — txn2/mcp-s3.
  • Composable library shape. Importable into custom AI gateways rather than only a standalone binary, enabling unified data-access architectures. Per GitHub — txn2/mcp-s3.

Connections 5

Outbound 5
scoped_to1
AI Runtime Infrastructure
extends1
Model Context Protocol (MCP)
integrates_with2
MinIOSeaweedFS
retrieves1
Object Storage

Resources 1

RepoHigh
github.com/txn2/mcp-s3

The source and docs for the open-source Go MCP server — secure defaults (read-only mode, GET size caps, prefix ACLs), composable library usage, and the tool contract for exposing S3 to agents.

Featured in

2026-06-27 The Storage Cost Inversion: When Object Storage Grew a Brain A 2026 NAND/flash shortage and a wave of cloud storage price hikes made fast storage scarce — and pushed AI memory onto S3. But object storage didn't just become the cheap tier. Under pressure it became the active substrate: an agentic data plane, an RL training buffer, and a hot KV-cache memory pool.