osrs-mcp
Combines screen vision, RuneLite HTTP API, click coordinates, OSRS Wiki lookups, and Grand Exchange prices to give AI agents eyes and knowledge for playing Old School RuneScape. Enables game state analysis, skill tracking, item/NPC lookup, and interactive automation.
README
osrs-mcp
MCP server for Old School RuneScape. Gives an AI agent eyes and knowledge for OSRS by combining:
- Screen vision — captures the RuneLite window and uses Claude or a local vision LLM to analyze game state
- RuneLite HTTP API — reads exact skill levels, inventory, and equipment directly from the game client (no vision needed)
- Click coordinates — resolves pixel positions for inventory slots, sidebar tabs, and UI regions
- OSRS Wiki — item, NPC, quest, and page lookups
- Grand Exchange prices — real-time prices, history, and profit calculator
- Skill training tools — XP calculations and items-to-level planning
Setup
cp .env.example .env
# Edit .env — set ANTHROPIC_API_KEY and optionally VISION_PROVIDER
npm install
npm run build
Vision provider
The vision tools support two backends, configured via .env:
Option A — Anthropic API (default)
ANTHROPIC_API_KEY=your-api-key-here
Option B — Local AI via LM Studio (or any Anthropic-compatible server)
Your local model must be vision/multimodal capable (e.g. qwen2-vl-7b-instruct, gemma-4, llava). Text-only models cannot analyze screenshots.
VISION_PROVIDER=local
LOCAL_AI_BASE_URL=http://localhost:1234
LOCAL_AI_API_KEY=lmstudio
LOCAL_AI_DETAIL_MODEL=your-vision-model-name
LOCAL_AI_FAST_MODEL=your-vision-model-name
LOCAL_AI_THINKING=0 # set to 1 if your model supports reasoning
RuneLite HTTP Server plugin
The get_stats, get_inventory, and get_equipment tools read live game data directly from RuneLite without using vision. To enable them:
- Open RuneLite → Plugin Hub → search HTTP Server → install and enable it
- The plugin exposes
http://localhost:8080/stats,/inv, and/equip - No extra configuration needed — the MCP connects on startup
These tools are more reliable and faster than their vision equivalents. Use them when you need ground-truth data (skill XP, exact item IDs, equipment slots).
macOS window capture
The screenshot tools use screencapture -l <CGWindowID> to capture just the RuneLite window. Window ID detection uses Swift + CGWindowListCopyWindowInfo — no extra tools required.
Fallback: If window ID detection fails, the full screen is captured. Vision analysis still works as long as RuneLite is visible.
Claude Desktop / Claude Code config
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"osrs": {
"command": "node",
"args": ["/absolute/path/to/osrs-mcp/dist/index.js"]
}
}
}
The server reads .env from the project root for ANTHROPIC_API_KEY / VISION_PROVIDER. You can also inline env vars in the config:
{
"mcpServers": {
"osrs": {
"command": "node",
"args": ["/absolute/path/to/osrs-mcp/dist/index.js"],
"env": { "ANTHROPIC_API_KEY": "sk-..." }
}
}
}
Or with tsx for development (no build step needed):
{
"mcpServers": {
"osrs": {
"command": "npx",
"args": ["tsx", "/absolute/path/to/osrs-mcp/src/index.ts"]
}
}
}
Tools
RuneLite HTTP API
Requires the HTTP Server plugin running on port 8080. Faster and more reliable than vision for these reads.
| Tool | Description |
|---|---|
get_stats |
All skill levels (base + boosted) and XP from /stats |
get_inventory |
Inventory items — item ID, quantity, slot 0-27 — from /inv |
get_equipment |
Equipped items by slot name (weapon, head, body…) from /equip |
Screen capture
| Tool | Description |
|---|---|
find_osrs_window |
Check if RuneLite is running and get its window ID |
capture_screen |
Take a screenshot of the RuneLite window |
Vision analysis
Uses Claude vision (Opus for detail, Haiku for fast reads). Falls back to local model if VISION_PROVIDER=local.
| Tool | Description |
|---|---|
analyze_screen |
Analyze the screen with vision (focus: inventory / stats / chat / minimap / full / bank / shop / npc) |
analyze_inventory |
List all items in the 28-slot inventory |
read_chat |
Extract all visible chat messages |
read_stats |
Read HP, Prayer, Run energy, Special attack from status bars |
read_minimap |
Read compass bearing and nearby entity counts |
get_game_state |
Full composite game state in one call |
detect_dialog |
Read NPC dialog, option menus, or level-up popups |
detect_combat_state |
Read combat indicators: XP drops, enemy HP bar, poison/venom, overhead prayers |
detect_loot |
Identify ground items visible after a kill |
Click coordinates
Returns pixel positions for UI elements. Layout is auto-detected from screenshot dimensions by default.
| Tool | Description |
|---|---|
detect_layout |
Detect whether RuneLite is in fixed (765×503) or resizable mode |
get_tab_coords |
Pixel coordinates to click a sidebar tab (prayer, inventory, skills, magic, combat_options, etc.) |
inventory_slot_coords |
Pixel center of an inventory slot 0–27 |
get_clickable_regions |
Bounding boxes for major UI panels: minimap, inventory, chat box, compass |
find_npc_on_screen |
Locate a named NPC in the current screenshot, returns {screen_x, screen_y, confidence} |
find_object_on_screen |
Locate an interactable object (bank chest, tree, ladder…), returns {screen_x, screen_y, confidence} |
Layout detection: fixed mode is identified by the exact dimensions 765×503 (or 1530×1006 on Retina). All other sizes are treated as resizable. Resizable coordinates are edge-anchored (inventory to bottom-right, chat to bottom-left) and computed from the actual captured dimensions.
Tab names (top row): combat_options · skills · quests · inventory · equipment · prayer · magic
Tab names (bottom row): clan_chat · friends · account_management · logout · settings · emotes · music
OSRS Wiki
| Tool | Description |
|---|---|
wiki_search |
Search the OSRS wiki |
wiki_page |
Get the intro/summary of a wiki page |
wiki_item_info |
Item details (alch values, members, weight, GE limit, etc.) |
wiki_npc_info |
NPC details (combat level, HP, attack styles, aggressive/poisonous) |
wiki_quest_info |
Quest details (difficulty, length, members, requirements, rewards) |
Grand Exchange
| Tool | Description |
|---|---|
ge_price |
Current GE price by item name or ID |
ge_price_history |
Price chart data (timestep: 5m / 1h / 6h / 24h) |
ge_mapping |
Full item name→ID mapping, filterable by name substring |
calculate_profit |
Flip profit calculator — applies 1% GE tax, returns per-item and total profit + ROI |
Utilities
| Tool | Description |
|---|---|
skill_xp_info |
XP required for a target level, or current level from an XP value |
calc_items_to_level |
How many actions/items to reach a target level given XP per action |
Development
npm run dev # Run directly via tsx (no build needed)
npm run build # Compile TypeScript → dist/
npm run inspect # Interactive tool tester via MCP Inspector
npm test # Run test suite (vitest)
npm run test:watch # Watch mode
Related
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.