DEV Community: Failed Ahmad

bluesky-mcp: Manage a Bluesky Account From Your AI Assistant

Failed Ahmad — Wed, 12 Aug 2026 22:08:05 +0000

What is this?

bluesky-mcp is a Model Context Protocol server for managing a Bluesky account from an AI assistant: post, reply, like, repost, follow/unfollow, search, and read your timeline and notifications.

It uses the official AT Protocol API with a handle + app password — no OAuth app, no approval process, and it works fine with 2FA enabled since it never touches your real account password.

This is the last of a four-part series — minecraft-mcp-server, streamlabs-mcp, and devto-mcp were covered in the previous three posts, all published (and promoted) using these exact tools.

Tools exposed

Tool	Does
`bluesky_get_timeline`	Get your home feed
`bluesky_get_author_feed`	Get a user's posts
`bluesky_search_posts`	Search public posts
`bluesky_get_post`	Get post details, incl. the `cid` needed to like/repost
`bluesky_get_profile`	Get profile info
`bluesky_get_followers` / `bluesky_get_following`	List a user's followers / follows
`bluesky_get_notifications` / `bluesky_mark_notifications_read`	Read/clear notifications
`bluesky_resolve_handle`	Look up a DID from a handle
`bluesky_post`	Post, optionally as a reply
`bluesky_delete_post`	Delete your own post
`bluesky_like` / `bluesky_unlike`	Like / undo a like
`bluesky_repost` / `bluesky_unrepost`	Repost / undo a repost
`bluesky_follow` / `bluesky_unfollow`	Follow / unfollow (by DID)

For the "un-" actions, pass the AT-URI of the record itself (the like/repost/follow record, returned in viewer.like / viewer.repost / viewer.following on posts and profiles) — not the original post or account's URI.

Getting started

git clone https://github.com/AhmadTariq1337/bluesky-mcp.git
cd bluesky-mcp
npm install
cp .env.example .env

Add your handle and an app password to .env (get one from Bluesky: Settings → Privacy and Security → App Passwords), then register it as an MCP server in your client's config — for Claude Desktop that's claude_desktop_config.json:

"bluesky": {
  "command": "node",
  "args": ["/absolute/path/to/bluesky-mcp/index.js"]
}

Fully quit and reopen your client (not just close the window) so it picks up the new server.

Bonus: standalone CLI

There's also a one-off post.js script for firing a single post from the terminal without going through an AI assistant:

node post.js "Just shipped a new project, check it out!"

Prints a link to the live post. 300-character limit, checked before sending.

Security

.env holds your real handle and app password, and is gitignored. App passwords are scoped and revocable from Bluesky's settings, unlike your main account password.

streamlabs-mcp: Let Your AI Assistant Run Streamlabs Desktop

Failed Ahmad — Wed, 12 Aug 2026 22:01:45 +0000

What is this?

streamlabs-mcp is a Model Context Protocol server that gives an AI assistant direct control over Streamlabs Desktop — switching scenes, showing/hiding sources, adjusting audio, starting/stopping streams or recordings, managing studio mode, and more.

It talks to Streamlabs Desktop's own local Remote Control API — the same one used by Streamlabs' official mobile remote app — so there's no screen automation or unofficial hacking involved.

Like the rest of this MCP series, it's not tied to one AI product. It's a plain stdio MCP server, so it works with any MCP-compatible client — Claude Desktop, Claude Code, Cursor, or anything else that can spawn a local server.

Disclaimer: this is an independent, community project — not made or endorsed by Streamlabs. Some tools (like toggling your live stream) can affect a real broadcast, so give your AI assistant clear boundaries about when it's allowed to use them.

What's in the box

59 tools across:

Scenes — list, get, switch, create, remove, rename, clear
Scene items — visibility, transform (position/scale/rotation/crop), flip, center, fit/stretch to screen
Sources — create, remove, rename, duplicate, refresh, settings, list by type
Audio — list sources, mute/unmute, set volume
Streaming & recording — status, toggle streaming, toggle recording, replay buffer, save replay
Studio mode — status, enable/disable, execute transition
Scene collections — list, get active, load, create, delete, rename
Notifications & performance — list/mark notifications, push notifications, CPU/FPS/dropped-frames stats

It's a pure stdio server — no cloud relay, no external service. Your assistant talks straight to Streamlabs Desktop over your own machine's loopback connection, and setup is just one token.

Getting started

git clone https://github.com/AhmadTariq1337/streamlabs-mcp.git
cd streamlabs-mcp
npm install

Grab your token from Streamlabs Desktop under Settings → Remote Control, then point your MCP client at node /path/to/streamlabs-mcp/index.js with STREAMLABS_TOKEN set in its environment. Full config examples for Claude Desktop, Claude Code, and Cursor are in the README.

Once it's running, try asking your assistant something like "list my Streamlabs scenes" or "what's my current streaming status?"

minecraft-mcp-server: Giving AI Agents a Body Inside Minecraft

Failed Ahmad — Wed, 12 Aug 2026 21:58:18 +0000

What is this?

minecraft-mcp-server is a Model Context Protocol server that puts an AI agent inside a real Minecraft Java Edition world through mineflayer. It exposes 61 tools covering movement, building, mining, combat, farming, inventory, villager trading, and world perception — so any MCP-compatible client can spawn a bot that actually sees the world, walks around it, and acts on it.

This isn't tied to any single AI product. It speaks plain MCP over stdio, so it works with any MCP host — Claude Desktop, Claude Code, Cursor, Cline, Windsurf, your own custom MCP client, or anything else that can launch a local MCP server. If your agent can talk MCP, it can play Minecraft.

Built for singleplayer worlds opened via "Open to LAN," offline-mode/cracked servers (e.g. TLauncher), and standard online-mode servers.

Tool chest — what your agent can do

Category	Tools
🔌 Connection	connect, disconnect, status
👀 Perception	position, nearby blocks, block lookup, find block, nearby entities, inventory
🚶 Movement	goto position, goto player, follow player, look at, look direction, stop, jump
🧱 Building	place block, dig block, build cuboid structure, clear area
🖐️ Interaction	chat, equip item, toss item, craft item, attack entity, activate block, eat, sleep, collect block (mine + auto-pickup)
🎒 Inventory	open/deposit/withdraw/close container, use furnace, give item (creative), move item slot
⚔️ Combat	sprint, sneak, use held item (block/draw/drink), shoot bow
🌾 Farming	till soil, plant seed, harvest crop
🌍 World	biome lookup, light level, get time, read sign
🗣️ Social	whisper, list players, open villager trades, execute trade
🎲 Misc	mount, dismount, respawn, wait

Every tool returns structured JSON and actionable error messages — no raw stack traces dumped on your agent.

Installation

git clone https://github.com/AhmadTariq1337/minecraft-mcp-server.git
cd minecraft-mcp-server
npm install
npm run build

This compiles TypeScript into dist/index.js — the entry point your MCP host will launch.

Configuring your MCP client

{
  "mcpServers": {
    "minecraft": {
      "command": "node",
      "args": ["/absolute/path/to/minecraft-mcp-server/dist/index.js"]
    }
  }
}

No API keys, no cloud dependency — everything runs locally over stdio.

Quick start in-game

Open your Minecraft world (singleplayer works great).
Press Escape → Open to LAN — note the port number shown in chat.
Ask your agent to connect: "Connect to localhost on port <port> as <bot-name>, auth offline."
That's it — your agent can now see the world, move, mine, build, and chat.

Version support

This rides on mineflayer, which determines actual protocol support. At time of writing, it supports up through Minecraft 1.21.x. Mojang's newer year.drop versioning isn't yet supported upstream — older clients like OptiFine 1.21.11 or vanilla 1.21.x work out of the box.

DEV Community: Failed Ahmad

bluesky-mcp: Manage a Bluesky Account From Your AI Assistant

What is this?

Tools exposed

Getting started

Bonus: standalone CLI

Security

Links

streamlabs-mcp: Let Your AI Assistant Run Streamlabs Desktop

What is this?

What's in the box

Getting started

Links

minecraft-mcp-server: Giving AI Agents a Body Inside Minecraft

What is this?

Tool chest — what your agent can do

Installation

Configuring your MCP client

Quick start in-game

Version support

Links