OSRS Wiki MCP

A local, read-only Model Context Protocol server for the Old School RuneScape Wiki. It gives AI clients ten focused tools for Wiki research, with validated structured output and source provenance on every successful response.

• Stateless: no player account, database, telemetry, or disk cache.
• Source-backed: results include canonical Wiki URLs and fetch timestamps; parsed pages also include an exact revision link.
• Bounded: response sizes, pagination, retries, and request time are capped.
• Read-only: the server only retrieves public Wiki data.

Quick start

Node.js 24 or newer and npx are required.

Codex

codex plugin marketplace add SSanderV/osrs-wiki-mcp --ref v1.1.3
codex plugin add osrs-wiki-mcp@osrs-wiki

Claude Code

claude plugin marketplace add SSanderV/[email protected] --scope user
claude plugin install osrs-wiki-mcp@osrs-wiki --scope user

Gemini CLI

gemini extensions install https://github.com/SSanderV/osrs-wiki-mcp --ref v1.1.3

The plugin and extension include the MCP setup plus a small Wiki-research skill. All three launch the pinned npm package [email protected].

Other MCP clients

Configure a standard-input/output server with:

{
  "command": "npx",
  "args": ["--yes", "[email protected]"]
}

Start a fresh client session after installation. If you previously configured an osrs-wiki MCP server directly, remove that registration first so the plugin is the only provider of the server.

Tools

| Tool | Inputs | Use it for | | --- | --- | --- | | search_wiki | query, optional limit and offset | Find canonical pages and snippets. | | get_wiki_page | title | Read a bounded, cleaned page with its section index. | | get_wiki_sections | title | List a page's sections. | | get_wiki_section | title, section | Read one section by its numeric index. | | get_item_info | item | Get an item's description, properties, bonuses, and creation facts. | | find_shop | item, optional limit and offset | Page through complete shop listings. | | find_drop_sources | item, optional limit and offset | Page through complete monster-drop listings. | | get_item_sources | item, optional perCategoryLimit | Get a bounded overview of drops, shops, recipes, and ground spawns. | | get_quest_requirements | quest | Get Wiki-sourced requirements without evaluating a player. | | get_monster_info | monster, optional variant | Get separate monster variants, map points, and access notes. |

Example requests:

• “What are the requirements for Dragon Slayer I?”
• “Where can an ironman obtain a rune scimitar?”
• “Find the exact Wiki section explaining Hespori mechanics.”

All public text inputs accept 1–256 Unicode characters. Tool schemas declare the exact limits and defaults. When results are truncated or upstream pagination is incomplete, warnings explains the safe follow-up action.

Responses and provenance

Successful calls return:

• content for clients that display readable text; and
• structuredContent, validated against the tool's MCP output schema.

Provenance includes the contributing Wiki URLs, original fetchedAt time, attribution, license details, and whether content was transformed. Parsed-page responses additionally include the exact revision ID and revision URL.

Failures use stable error codes such as NOT_FOUND, UPSTREAM_TIMEOUT, UPSTREAM_RATE_LIMITED, UPSTREAM_UNAVAILABLE, UPSTREAM_INVALID_RESPONSE, and RESPONSE_TOO_LARGE. Invalid arguments remain standard JSON-RPC parameter errors.

Reliability and privacy

• A tool call has a 30-second total budget; each upstream request gets at most 20 seconds within that budget.
• Transient network and rate-limit failures receive at most two bounded retries, respecting Retry-After only when time remains.
• Responses are capped at 5 MiB. Large Bucket scans stop after 10,000 raw rows and report incomplete or truncated results.
• Valid upstream responses may be cached in memory for five minutes. The cache is limited per process, never written to disk, and cleared on restart.
• Standard output is reserved for MCP traffic. Sanitized diagnostics use standard error.

Scope

The server deliberately does not provide player progress, hiscores, Grand Exchange prices, DPS calculations, Wiki images, persistent storage, hosting, or a local Wiki mirror. Quest requirements are facts rather than met/missing evaluations, and monster variants are never combined.

Development

npm.cmd ci --ignore-scripts
npm.cmd run typecheck
npm.cmd test
npm.cmd run build
npm.cmd run smoke:stdio
npm.cmd run pack:check

Tests use synthetic fixtures and do not contact the live Wiki. The optional live smoke makes two low-volume requests and does not persist responses:

npm.cmd run test:live

See CONTRIBUTING.md for development guidance and SECURITY.md for private vulnerability reporting.

Licensing and attribution

The source code is MIT licensed. Retrieved Old School RuneScape Wiki content remains subject to CC BY-NC-SA 3.0, including its non-commercial and ShareAlike conditions. Downstream users are responsible for how they reuse that content. See THIRD_PARTY_NOTICES.md.

RuneScape and Old School RuneScape are trademarks of Jagex Limited. This independent project is not affiliated with, endorsed by, or sponsored by Jagex Limited, Weird Gloop, or the Old School RuneScape Wiki.