Torn MCP Server
Remote MCP server for the Torn City API v2, enabling querying of player, faction, market, and other game data, plus aggregated intelligence reports, all via natural language.
README
Torn MCP Server
A remote Model Context Protocol server for the Torn City API v2, running on Cloudflare Workers. Connect from VS Code (or any MCP client) anywhere — no local install. You supply your Torn API key via the X-Torn-Api-Key header; it is never stored, logged, or shown to the model.
Tools: one grouped tool per Torn tag (torn_user, torn_faction, torn_torn, torn_company, torn_market, torn_racing, torn_forum, torn_property, torn_key) covering all 205 API paths via an endpoint argument — plus 12 intelligence tools that aggregate endpoints into structured summaries (analyze_player, war_readiness_report, find_profitable_items, …) and torn_list_endpoints for discovery.
[!NOTE] Read-only. The Torn API v2 exposes only
GETendpoints — this server can never modify your account or take in-game actions.
Install
Live endpoint (hosted instance):
https://torn-mcp.yoonutz.workers.dev/mcp
One-click install:
Click a badge → the client opens with the server pre-filled. It installs with a YOUR_TORN_API_KEY placeholder — replace it with your real key (get one) in the client's MCP settings after install.
Setup guides (these clients have no one-click protocol — badge links to their MCP docs; use the config blocks below):
Or set it up manually — pick your client below. Prefer to self-host? See Deploy.
<details> <summary><b>Claude Code</b> (CLI)</summary>
claude mcp add --transport http torn https://torn-mcp.yoonutz.workers.dev/mcp \
--header "X-Torn-Api-Key: YOUR_TORN_API_KEY"
Check it: claude mcp list. Remove: claude mcp remove torn.
</details>
<details> <summary><b>VS Code</b> (CLI)</summary>
macOS / Linux (bash, zsh):
code --add-mcp '{"name":"torn","type":"http","url":"https://torn-mcp.yoonutz.workers.dev/mcp","headers":{"X-Torn-Api-Key":"YOUR_TORN_API_KEY"}}'
Windows PowerShell — escape the inner quotes with \" (the code.cmd shim strips plain quotes otherwise):
code --add-mcp '{\"name\":\"torn\",\"type\":\"http\",\"url\":\"https://torn-mcp.yoonutz.workers.dev/mcp\",\"headers\":{\"X-Torn-Api-Key\":\"YOUR_TORN_API_KEY\"}}'
Insiders: use code-insiders. Then open Copilot Chat → Agent mode → 🛠️ tools and enable torn. (If quoting still fights you, use the manual mcp.json below — no escaping needed.)
Manual alternative — user mcp.json (Command Palette → MCP: Open User Configuration):
{
"servers": {
"torn": {
"type": "http",
"url": "https://torn-mcp.yoonutz.workers.dev/mcp",
"headers": { "X-Torn-Api-Key": "YOUR_TORN_API_KEY" }
}
}
}
</details>
<details> <summary><b>Cursor</b></summary>
Edit ~/.cursor/mcp.json (global) or .cursor/mcp.json (project):
{
"mcpServers": {
"torn": {
"url": "https://torn-mcp.yoonutz.workers.dev/mcp",
"headers": { "X-Torn-Api-Key": "YOUR_TORN_API_KEY" }
}
}
}
</details>
<details> <summary><b>Windsurf</b></summary>
Edit ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"torn": {
"serverUrl": "https://torn-mcp.yoonutz.workers.dev/mcp",
"headers": { "X-Torn-Api-Key": "YOUR_TORN_API_KEY" }
}
}
}
</details>
<details> <summary><b>Claude Desktop</b></summary>
Settings → Developer → Edit Config (claude_desktop_config.json). Recent versions accept a remote URL directly:
{
"mcpServers": {
"torn": {
"url": "https://torn-mcp.yoonutz.workers.dev/mcp",
"headers": { "X-Torn-Api-Key": "YOUR_TORN_API_KEY" }
}
}
}
If your version only supports stdio servers (the connection fails), use the mcp-remote bridge instead:
{
"mcpServers": {
"torn": {
"command": "npx",
"args": [
"mcp-remote",
"https://torn-mcp.yoonutz.workers.dev/mcp",
"--header",
"X-Torn-Api-Key:YOUR_TORN_API_KEY"
]
}
}
}
Restart Claude Desktop.
</details>
<details> <summary><b>Visual Studio</b> (2022 17.14+)</summary>
Add a .mcp.json to your solution or %USERPROFILE%\.mcp.json:
{
"servers": {
"torn": {
"type": "http",
"url": "https://torn-mcp.yoonutz.workers.dev/mcp",
"headers": { "X-Torn-Api-Key": "YOUR_TORN_API_KEY" }
}
}
}
</details>
<details> <summary><b>Continue</b></summary>
Edit ~/.continue/config.yaml:
mcpServers:
- name: torn
type: streamable-http
url: https://torn-mcp.yoonutz.workers.dev/mcp
requestOptions:
headers:
X-Torn-Api-Key: YOUR_TORN_API_KEY
</details>
The Claude Code and VS Code CLI commands above are tested. The other clients use the same endpoint +
X-Torn-Api-Keyheader (verified working), but their config key names and file paths can change between versions — check the client's own MCP docs if a connection fails.
Security: putting the key inline lands it in config files / shell history. Where the client supports it (e.g. VS Code
${input:...}prompts), prefer that over a plaintext key.
Get a Torn API Key
Torn → Settings → API Keys → create a key. A Limited or Minimal key covers most tools; faction/market selections may need broader scope. Paste it into the X-Torn-Api-Key header.
Tools
One grouped tool per Torn tag, each covering all of that tag's endpoints:
| Tool | Endpoints | Example endpoint values |
|---|---|---|
torn_user |
64 | profile, battlestats, bars, cooldowns, money, events |
torn_faction |
34 | basic, members, wars, attacks, chain |
torn_torn |
26 | items, stats, timestamp, territory |
torn_company |
10 | profile, employees, stock, news |
torn_market |
8 | itemmarket, bazaar, auctionhouse |
torn_racing |
8 | races, cars, tracks, records |
torn_forum |
6 | categories, threads, posts |
torn_property |
2 | property |
torn_key |
2 | info, log |
torn_list_endpoints |
— | discovery: lists every endpoint per tag |
Each tool takes:
endpoint(required) — which data type to fetch (full list per tool, or calltorn_list_endpoints).id(optional) — entity id; used when the endpoint is entity-scoped or requires one.params(optional) — extra query options (limit,from,to,sort,cat, …).
The Torn key is not a tool parameter — it comes from the request header, so it never enters model context or client transcripts.
Intelligence tools
Higher-level tools that aggregate multiple endpoints and return structured summaries instead:
| Tool | Aggregates | Returns |
|---|---|---|
analyze_player |
user/profile + personalstats | status, activity, life, social, stats |
summarize_player |
user/profile | condensed one-glance snapshot |
compare_players |
user/profile ×N | side-by-side + level gap to top |
summarize_faction |
faction/basic + members | counts by position and activity |
faction_member_activity |
faction/members | online / idle / offline buckets |
war_readiness_report |
faction/members | availability-based readiness score |
territory_summary |
faction + torn territory | holdings + global sample |
crime_analysis |
faction/crimes | status/difficulty counts, success rate |
summarize_company |
company/profile + employees | profile + headcount |
item_market_analysis |
market/itemmarket + torn/items | depth, price band, market value |
market_analysis |
market/itemmarket ×N | items ranked by spread |
find_profitable_items |
market/itemmarket + torn/items ×N | items ranked by margin |
War readiness is availability-based (okay/hospital/traveling, online, on-wall, in-OC) — per-member battlestats are not exposed by the Torn API.
Security & privacy
- Key supplied via
X-Torn-Api-Keyheader only — never a tool parameter. - Never stored, never logged, never returned in error messages.
- Upstream calls are pinned to
https://api.torn.com(SSRF guard) withUser-Agent: torn-mcp. - Per-key rate limiting (~100 req/min, Torn's cap) via a Durable Object — returns a clear error instead of hammering Torn.
Dependency advisories
Dev/build tooling (vitest, vite, esbuild) and a transitive (jsondiffpatch) are pinned to patched versions. The remaining Dependabot alerts live in the agents SDK chain (@ai-sdk/provider-utils, ai, and agents' own AI-Playground / OAuth-callback / email-routing advisories). None are reachable here — this server exposes no playground, OAuth flow, or email routing — and the only offered upgrade (agents@0.16) forces a zod 3→4 major that conflicts with the MCP SDK. They are tracked, not exploitable, and will be cleared when agents ships a zod-3-compatible patch.
Scope & roadmap
Covered: all 9 Torn tags (160 endpoints / all 205 paths) as raw JSON, plus 12 intelligence tools and a discovery tool. Still on the roadmap from the retired Fastify/Docker design (superseded spec): MCP resources, prompts, and richer per-endpoint output typing.
License
MIT — // @license MIT headers are included in all source files.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.