feat(mcp): add MCP plugin packages for DocSearch
Backport the mcp/ folder from v5 so plugin configs for Cursor, Claude, ChatGPT, and Codex are available on main.
This commit is contained in:
parent
3f2b6e2fdf
commit
9378e83941
20 changed files with 608 additions and 0 deletions
9
mcp/README.md
Normal file
9
mcp/README.md
Normal file
|
|
@ -0,0 +1,9 @@
|
|||
# DocSearch MCP
|
||||
|
||||
Client plugin packages for the public Algolia DocSearch MCP endpoint live in [`plugins/docsearch`](plugins/docsearch).
|
||||
|
||||
```text
|
||||
https://mcp.algolia.com/1/docsearch/mcp
|
||||
```
|
||||
|
||||
The plugins connect AI coding clients to the public DocSearch documentation corpus. They do not expose crawler operations, DocSearch administration, private documentation, or Algolia index credentials.
|
||||
72
mcp/plugins/docsearch/README.md
Normal file
72
mcp/plugins/docsearch/README.md
Normal file
|
|
@ -0,0 +1,72 @@
|
|||
# Algolia DocSearch MCP Plugins
|
||||
|
||||
This directory contains client plugin packages for the public Algolia DocSearch MCP endpoint:
|
||||
|
||||
```text
|
||||
https://mcp.algolia.com/1/docsearch/mcp
|
||||
```
|
||||
|
||||
The plugins connect AI coding clients to the public DocSearch documentation corpus. They do not expose crawler operations, DocSearch administration, private documentation, or Algolia index credentials.
|
||||
|
||||
## Packages
|
||||
|
||||
- `mcp/plugins/docsearch/chatgpt/algolia-docsearch` - Universal ChatGPT and Codex plugin package with manifest, MCP config, skill, assets, and submission materials.
|
||||
- `mcp/plugins/docsearch/cursor/algolia-docsearch` - Cursor plugin package with MCP config, rule, skill, and README.
|
||||
- `mcp/plugins/docsearch/claude/algolia-docsearch` - Claude Code plugin package with MCP config, skill, command, and README.
|
||||
|
||||
## Public MCP Tools
|
||||
|
||||
You can query in natural language — full sentences and questions work well.
|
||||
|
||||
### algolia_docsearch_search_docs
|
||||
|
||||
One-shot lookup, and the right default for most questions. Pass a `library` and a `query`; it resolves the best matching documentation set and returns ranked content in a single call:
|
||||
|
||||
```json
|
||||
{
|
||||
"library": "Next.js",
|
||||
"query": "how do middleware matchers work"
|
||||
}
|
||||
```
|
||||
|
||||
If the library is ambiguous, it returns candidate documentation sets instead — pick a `docset_id` and pass it to `algolia_docsearch_query_docs`.
|
||||
|
||||
### algolia_docsearch_resolve_docset
|
||||
|
||||
Step 1 of the manual flow. Finds matching documentation sets and returns candidates with:
|
||||
|
||||
- `docset_id`
|
||||
- title and description
|
||||
- optional quality signals such as `trustScore`, `benchmarkScore`, and `popularityScore`
|
||||
|
||||
```json
|
||||
{
|
||||
"query": "Next.js app router"
|
||||
}
|
||||
```
|
||||
|
||||
### algolia_docsearch_query_docs
|
||||
|
||||
Step 2 of the manual flow. Queries documentation content for the selected `docset_id` values. Pass several when a question spans multiple products:
|
||||
|
||||
```json
|
||||
{
|
||||
"query": "middleware matcher config",
|
||||
"docsetIds": ["nextjs"]
|
||||
}
|
||||
```
|
||||
|
||||
## Claude Desktop And claude.ai
|
||||
|
||||
Claude Desktop and claude.ai can connect directly to the same remote MCP endpoint when remote connectors are available in the client:
|
||||
|
||||
```json
|
||||
{
|
||||
"algolia-docsearch": {
|
||||
"type": "http",
|
||||
"url": "https://mcp.algolia.com/1/docsearch/mcp"
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
No user authentication is required for the public DocSearch MCP endpoint.
|
||||
|
|
@ -0,0 +1,44 @@
|
|||
{
|
||||
"name": "algolia-docsearch",
|
||||
"version": "1.0.0",
|
||||
"description": "Search current public developer documentation from the Algolia DocSearch corpus.",
|
||||
"author": {
|
||||
"name": "Algolia",
|
||||
"url": "https://www.algolia.com"
|
||||
},
|
||||
"homepage": "https://docsearch.algolia.com/docs/mcp/overview",
|
||||
"repository": "https://github.com/algolia/docsearch",
|
||||
"license": "MIT",
|
||||
"keywords": [
|
||||
"algolia",
|
||||
"developer-docs",
|
||||
"documentation",
|
||||
"docsearch",
|
||||
"mcp"
|
||||
],
|
||||
"skills": "./skills/",
|
||||
"mcpServers": "./.mcp.json",
|
||||
"interface": {
|
||||
"displayName": "Algolia DocSearch",
|
||||
"shortDescription": "Current public developer documentation from the DocSearch corpus",
|
||||
"longDescription": "Search current public documentation for libraries, frameworks, SDKs, APIs, CLIs, and cloud services. DocSearch resolves the relevant documentation set, returns ranked excerpts, and provides source URLs for grounded answers.",
|
||||
"developerName": "Algolia",
|
||||
"category": "Developer Tools",
|
||||
"capabilities": [
|
||||
"Interactive",
|
||||
"Read"
|
||||
],
|
||||
"websiteURL": "https://docsearch.algolia.com/docs/mcp/overview",
|
||||
"privacyPolicyURL": "https://www.algolia.com/policies/privacy",
|
||||
"termsOfServiceURL": "https://www.algolia.com/policies/terms",
|
||||
"defaultPrompt": [
|
||||
"Explain middleware matchers from the current Next.js docs.",
|
||||
"Show how React's useTransition hook works.",
|
||||
"Find Algolia guidance for searchable attributes."
|
||||
],
|
||||
"brandColor": "#5468FF",
|
||||
"composerIcon": "./assets/docsearch-monogram.svg",
|
||||
"logo": "./assets/docsearch-monogram.svg",
|
||||
"screenshots": []
|
||||
}
|
||||
}
|
||||
1
mcp/plugins/docsearch/chatgpt/algolia-docsearch/.gitignore
vendored
Normal file
1
mcp/plugins/docsearch/chatgpt/algolia-docsearch/.gitignore
vendored
Normal file
|
|
@ -0,0 +1 @@
|
|||
SUBMISSION.md
|
||||
|
|
@ -0,0 +1,8 @@
|
|||
{
|
||||
"mcpServers": {
|
||||
"algolia-docsearch": {
|
||||
"type": "http",
|
||||
"url": "https://mcp.algolia.com/1/docsearch/mcp"
|
||||
}
|
||||
}
|
||||
}
|
||||
57
mcp/plugins/docsearch/chatgpt/algolia-docsearch/README.md
Normal file
57
mcp/plugins/docsearch/chatgpt/algolia-docsearch/README.md
Normal file
|
|
@ -0,0 +1,57 @@
|
|||
# Algolia DocSearch Plugin for ChatGPT and Codex
|
||||
|
||||
Algolia DocSearch connects ChatGPT and Codex to the public DocSearch MCP endpoint for current public developer documentation.
|
||||
|
||||
## What's included
|
||||
|
||||
- **Plugin manifest** - Identifies the universal ChatGPT and Codex plugin and its listing metadata.
|
||||
- **MCP server** - Connects to `https://mcp.algolia.com/1/docsearch/mcp` without user authentication.
|
||||
- **Skill** - Routes public developer documentation questions through DocSearch and requires source-grounded answers.
|
||||
|
||||
The plugin teaches the one-shot `search_docs` lookup and the `resolve_docset` plus `query_docs` flow.
|
||||
|
||||
## Public submission
|
||||
|
||||
Submit this as **With MCP** using the universal MCP URL:
|
||||
|
||||
```text
|
||||
https://mcp.algolia.com/1/docsearch/mcp
|
||||
```
|
||||
|
||||
Do not submit an existing ChatGPT integration ID. The OpenAI submission portal scans the production MCP URL directly. See [SUBMISSION.md](SUBMISSION.md) for listing copy, tool annotation requirements, starter prompts, test cases, and the pre-submission checklist.
|
||||
|
||||
## Local ChatGPT testing
|
||||
|
||||
1. Enable Developer mode in ChatGPT under **Settings → Security and login**.
|
||||
2. Open ChatGPT Plugins, add a connection, and enter the MCP URL above.
|
||||
3. Test each tool and the bundled skill in a new chat.
|
||||
|
||||
If a packaged local plugin must reference the registered ChatGPT connection, copy its technical ID from the browser URL and add a `.app.json` file:
|
||||
|
||||
```json
|
||||
{
|
||||
"apps": {
|
||||
"algolia-docsearch": {
|
||||
"id": "plugin_asdk_app_<registered-id>"
|
||||
}
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
Then add `"apps": "./.app.json"` to `.codex-plugin/plugin.json`. The committed package intentionally omits this machine-specific development ID.
|
||||
|
||||
## Available tools
|
||||
|
||||
### `algolia_docsearch_search_docs`
|
||||
|
||||
The default one-shot lookup. It resolves a documentation set from `library`, searches it with `query`, and returns ranked documentation chunks with source URLs.
|
||||
|
||||
### `algolia_docsearch_resolve_docset`
|
||||
|
||||
Finds candidate documentation sets for an official product, library, SDK, or platform.
|
||||
|
||||
### `algolia_docsearch_query_docs`
|
||||
|
||||
Queries one or more selected `docset_id` values. Use it after resolving an ambiguous library or when a question spans products.
|
||||
|
||||
This plugin is limited to public developer documentation. It does not expose crawler operations, DocSearch administration, private documentation, or Algolia index credentials.
|
||||
|
|
@ -0,0 +1,4 @@
|
|||
<svg xmlns="http://www.w3.org/2000/svg" viewBox="-10 -10 96 98">
|
||||
<path fill="#5468ff" d="M36.493 77.289H3.074A3.067 3.067 0 0 1 0 74.228h36.523c12.769.062 24.6-6.663 31.038-17.642a35.45 35.45 0 0 0 .174-35.584C61.404 9.962 49.639 3.122 36.87 3.061H0A3.067 3.067 0 0 1 3.074 0H36.87c10.31.01 20.193 4.1 27.473 11.369a38.6 38.6 0 0 1 11.341 27.384c-.06 21.366-17.741 38.536-39.19 38.536z"/>
|
||||
<path fill="#5468ff" d="M0 69.045h23.711a11.78 11.78 0 0 0 7.933-3.031H0zM0 60.8h36.374c.734-.987 1.418-1.975 2.053-3.07H0zm0-8.254h41.025a30 30 0 0 0 1.091-3.06H0zm0-8.244h43.306c.149-.987.268-2.034.337-3.061H0zm0-8.244h43.633a30 30 0 0 1-.337-3.071H0zm0-8.254h42.116a36 36 0 0 0-1.09-3.061H0zm0-8.245h38.427a33 33 0 0 1-2.053-3.06H0zM0 8.244v3.071h31.674a19.04 19.04 0 0 0-7.934-3.031z"/>
|
||||
</svg>
|
||||
|
After Width: | Height: | Size: 791 B |
|
|
@ -0,0 +1,51 @@
|
|||
---
|
||||
name: algolia-docsearch-mcp
|
||||
description: Use when the user asks about public developer documentation for a library, framework, SDK, API, CLI tool, or cloud service. Fetch current docs from Algolia DocSearch MCP and cite source URLs.
|
||||
---
|
||||
|
||||
Use Algolia DocSearch MCP when the answer depends on current public developer documentation.
|
||||
|
||||
## Scope
|
||||
|
||||
Use this skill for:
|
||||
|
||||
- Setup and configuration questions about public libraries, frameworks, SDKs, APIs, CLI tools, and cloud services.
|
||||
- API reference details, migrations, and code examples that should match current documentation.
|
||||
- Questions that ask for documentation sources or citations.
|
||||
|
||||
Do not use this skill for:
|
||||
|
||||
- Private or internal documentation.
|
||||
- Local code review or business logic.
|
||||
- General programming concepts that do not require external documentation.
|
||||
- Questions already answered by documentation the user supplied.
|
||||
- DocSearch crawler operations, administration, or private Algolia indices.
|
||||
|
||||
## Default flow
|
||||
|
||||
For most questions, call `algolia_docsearch_search_docs`.
|
||||
|
||||
1. Set `library` to the official product, library, SDK, or platform name.
|
||||
2. Set `query` to the actual documentation question in natural language.
|
||||
3. Answer only from the returned documentation chunks.
|
||||
4. Include returned source URLs when available.
|
||||
|
||||
If the selected documentation set is ambiguous or irrelevant, do not answer from it. Retry with the exact official product name or use the manual flow.
|
||||
|
||||
## Manual flow
|
||||
|
||||
Use `algolia_docsearch_resolve_docset` followed by `algolia_docsearch_query_docs` when:
|
||||
|
||||
- The question spans multiple products.
|
||||
- The right documentation set is ambiguous.
|
||||
- The user asks to compare guidance across products.
|
||||
|
||||
Resolve the documentation set once, select the official vendor's best match, and pass its `docset_id` to `algolia_docsearch_query_docs`. Pass multiple IDs when the question genuinely spans products.
|
||||
|
||||
## Answering
|
||||
|
||||
- Ground the answer only in returned documentation content.
|
||||
- Prefer official vendor documentation sets over third-party mentions.
|
||||
- Include source links next to the claims they support.
|
||||
- If no relevant documentation is returned, say so and ask for a more specific product name or documentation source.
|
||||
- Never invent citations, credentials, private data, or unsupported capabilities.
|
||||
|
|
@ -0,0 +1,15 @@
|
|||
{
|
||||
"name": "algolia-docsearch-marketplace",
|
||||
"description": "Algolia DocSearch plugins for Claude Code — search up-to-date public developer documentation from the DocSearch corpus.",
|
||||
"owner": {
|
||||
"name": "Algolia"
|
||||
},
|
||||
"plugins": [
|
||||
{
|
||||
"name": "algolia-docsearch",
|
||||
"source": ".",
|
||||
"description": "up-to-date public developer documentation from the DocSearch corpus.",
|
||||
"version": "1.0.0"
|
||||
}
|
||||
]
|
||||
}
|
||||
|
|
@ -0,0 +1,8 @@
|
|||
{
|
||||
"name": "algolia-docsearch",
|
||||
"description": "Algolia DocSearch MCP plugin for searching up-to-date public developer documentation from the DocSearch corpus.",
|
||||
"version": "1.0.0",
|
||||
"author": {
|
||||
"name": "Algolia"
|
||||
}
|
||||
}
|
||||
6
mcp/plugins/docsearch/claude/algolia-docsearch/.mcp.json
Normal file
6
mcp/plugins/docsearch/claude/algolia-docsearch/.mcp.json
Normal file
|
|
@ -0,0 +1,6 @@
|
|||
{
|
||||
"algolia-docsearch": {
|
||||
"type": "http",
|
||||
"url": "https://mcp.algolia.com/1/docsearch/mcp"
|
||||
}
|
||||
}
|
||||
63
mcp/plugins/docsearch/claude/algolia-docsearch/README.md
Normal file
63
mcp/plugins/docsearch/claude/algolia-docsearch/README.md
Normal file
|
|
@ -0,0 +1,63 @@
|
|||
# Algolia DocSearch Plugin for Claude Code
|
||||
|
||||
Algolia DocSearch connects Claude Code to the public DocSearch MCP endpoint for current public developer documentation.
|
||||
|
||||
## What's Included
|
||||
|
||||
- **MCP Server** - Connects Claude Code to `https://mcp.algolia.com/1/docsearch/mcp`.
|
||||
- **Skill** - Auto-triggers documentation lookups for public developer docs questions.
|
||||
- **Command** - `/algolia-docsearch:docs` for manual documentation queries.
|
||||
|
||||
The plugin teaches the one-shot `search_docs` lookup and the `resolve_docset` + `query_docs` flow.
|
||||
|
||||
## Installation
|
||||
|
||||
For local beta testing, install this plugin package from:
|
||||
|
||||
```text
|
||||
mcp/plugins/docsearch/claude/algolia-docsearch
|
||||
```
|
||||
|
||||
The MCP server config is also available at:
|
||||
|
||||
```text
|
||||
mcp/plugins/docsearch/claude/algolia-docsearch/.mcp.json
|
||||
```
|
||||
|
||||
For Claude Code marketplace install from this repository:
|
||||
|
||||
```text
|
||||
/plugin marketplace add algolia/docsearch
|
||||
/plugin install algolia-docsearch@algolia-docsearch-marketplace
|
||||
```
|
||||
|
||||
## Available Tools
|
||||
|
||||
### algolia_docsearch_search_docs
|
||||
|
||||
One-shot lookup, and the right default for most questions. Pass a `library` (product or platform) and a `query` (the question); it resolves the best documentation set and returns ranked content in a single call.
|
||||
|
||||
### algolia_docsearch_resolve_docset
|
||||
|
||||
Step 1 of the manual flow. Finds matching documentation sets and returns candidates with a `docset_id`, title, description, and ranking signals.
|
||||
|
||||
### algolia_docsearch_query_docs
|
||||
|
||||
Step 2 of the manual flow. Queries documentation content for the `docset_id` values returned by `algolia_docsearch_resolve_docset`. Pass several `docsetIds` when a question spans multiple products.
|
||||
|
||||
## Usage Examples
|
||||
|
||||
Ask Claude questions like:
|
||||
|
||||
- "How do I configure middleware in Next.js?"
|
||||
- "Show me Stripe webhook signature verification from the docs."
|
||||
- "What is the current Algolia InstantSearch React setup?"
|
||||
|
||||
Or use the manual command:
|
||||
|
||||
```text
|
||||
/algolia-docsearch:docs Next.js middleware matcher
|
||||
/algolia-docsearch:docs Stripe webhook signature verification
|
||||
```
|
||||
|
||||
The plugin is for public developer documentation only. It does not expose crawler operations, DocSearch administration, private docs, or Algolia index credentials.
|
||||
|
|
@ -0,0 +1,31 @@
|
|||
---
|
||||
description: Look up public developer documentation with Algolia DocSearch
|
||||
argument-hint: [query]
|
||||
---
|
||||
|
||||
# /algolia-docsearch:docs
|
||||
|
||||
Fetch current public developer documentation from the Algolia DocSearch MCP corpus.
|
||||
|
||||
## Usage
|
||||
|
||||
```text
|
||||
/algolia-docsearch:docs <library-or-product> [topic]
|
||||
```
|
||||
|
||||
For most lookups, call `algolia_docsearch_search_docs` in a single step:
|
||||
|
||||
1. Set `library` to the product, library, or platform (use the official name).
|
||||
2. Set `query` to the topic or question, in natural language.
|
||||
3. Answer using the returned docs and include source URLs when available.
|
||||
|
||||
If the library is ambiguous, `algolia_docsearch_search_docs` returns candidate documentation sets — pick the right `docset_id` and pass it to `algolia_docsearch_query_docs`. Use the `algolia_docsearch_resolve_docset` + `algolia_docsearch_query_docs` flow when a question spans several products.
|
||||
|
||||
## Examples
|
||||
|
||||
```text
|
||||
/algolia-docsearch:docs Next.js middleware matcher
|
||||
/algolia-docsearch:docs Stripe webhook signature verification
|
||||
/algolia-docsearch:docs Algolia InstantSearch React configure search client
|
||||
/algolia-docsearch:docs React Server Components data fetching
|
||||
```
|
||||
|
|
@ -0,0 +1,72 @@
|
|||
---
|
||||
name: algolia-docsearch-mcp
|
||||
description: Use this skill when the user asks about public developer documentation, SDKs, APIs, libraries, frameworks, setup, configuration, or code examples. Fetch current docs from Algolia DocSearch MCP and cite source URLs.
|
||||
---
|
||||
|
||||
When the user asks about public developer documentation, use Algolia DocSearch MCP instead of relying only on training data.
|
||||
|
||||
## When To Use
|
||||
|
||||
Use this skill when the user:
|
||||
|
||||
- Asks setup or configuration questions for a public library, framework, SDK, API, or developer tool.
|
||||
- Requests code examples that should match current docs.
|
||||
- Needs API reference details, migration guidance, or command syntax.
|
||||
- Mentions a library or framework whose docs may have changed recently.
|
||||
- Asks for sources or citations from public documentation.
|
||||
|
||||
Do not use this skill for private company docs, internal repositories, unpublished APIs, or questions where the user already supplied the relevant docs.
|
||||
|
||||
## Tools
|
||||
|
||||
DocSearch MCP exposes three tools. You can query in natural language — full sentences and questions work well.
|
||||
|
||||
- `algolia_docsearch_search_docs` — one-shot lookup. Use this by default.
|
||||
- `algolia_docsearch_resolve_docset` — step 1 of the manual flow: find documentation sets.
|
||||
- `algolia_docsearch_query_docs` — step 2 of the manual flow: fetch content for chosen docsets.
|
||||
|
||||
## Default Flow: One-Shot Search
|
||||
|
||||
Call `algolia_docsearch_search_docs` for most lookups.
|
||||
|
||||
Input:
|
||||
|
||||
- `library`: the product, library, SDK, or platform. Use the official name (for example: `Next.js`, `Stripe`, `Algolia InstantSearch`).
|
||||
- `query`: the question, in natural language (for example: `how do middleware matchers work`).
|
||||
|
||||
It resolves the best matching documentation set and returns ranked content in a single call. If `library` is ambiguous, it returns candidate documentation sets instead — pick the right one and pass its `docset_id` to `algolia_docsearch_query_docs`, or retry with the official product name.
|
||||
|
||||
## Manual Flow: Resolve, Then Query
|
||||
|
||||
Use the two-step flow when a question spans several products, or when you want to inspect and hand-pick documentation sets.
|
||||
|
||||
### Step 1: Resolve The Docset
|
||||
|
||||
Call `algolia_docsearch_resolve_docset`.
|
||||
|
||||
Input:
|
||||
|
||||
- `query`: describe the product, library, SDK, or platform. Including the vendor name improves matching.
|
||||
- `topN`: optional. Use the default unless the initial result is ambiguous.
|
||||
|
||||
Each candidate returns a `title`, `docset_id`, `description`, and ranking signals (`trustScore`, `benchmarkScore`, `popularityScore`). Prefer the official vendor's docset over third-party mentions, breaking ties by the higher scores.
|
||||
|
||||
### Step 2: Query The Docs
|
||||
|
||||
Call `algolia_docsearch_query_docs`.
|
||||
|
||||
Input:
|
||||
|
||||
- `query`: what to find in the docs, in natural language.
|
||||
- `docsetIds`: an array of the `docset_id` values you selected. Pass several when a question spans multiple products.
|
||||
|
||||
## Answer With Sources
|
||||
|
||||
Use the returned documentation content to answer the user directly.
|
||||
|
||||
Guidelines:
|
||||
|
||||
- Include source URLs from the tool result when available.
|
||||
- Say when DocSearch does not return a relevant result.
|
||||
- Do not invent source citations.
|
||||
- Keep the answer focused on the user's question.
|
||||
|
|
@ -0,0 +1,17 @@
|
|||
{
|
||||
"name": "algolia-docsearch",
|
||||
"version": "1.0.0",
|
||||
"description": "Algolia DocSearch MCP plugin for searching up-to-date public developer documentation from the DocSearch corpus.",
|
||||
"author": {
|
||||
"name": "Algolia"
|
||||
},
|
||||
"homepage": "https://docsearch.algolia.com/docs/mcp/overview",
|
||||
"keywords": [
|
||||
"documentation",
|
||||
"mcp",
|
||||
"algolia",
|
||||
"docsearch",
|
||||
"developer-docs"
|
||||
],
|
||||
"logo": "assets/docsearch-monogram.svg"
|
||||
}
|
||||
47
mcp/plugins/docsearch/cursor/algolia-docsearch/README.md
Normal file
47
mcp/plugins/docsearch/cursor/algolia-docsearch/README.md
Normal file
|
|
@ -0,0 +1,47 @@
|
|||
# Algolia DocSearch Plugin for Cursor
|
||||
|
||||
Algolia DocSearch connects Cursor to the public DocSearch MCP endpoint for current public developer documentation.
|
||||
|
||||
## What's Included
|
||||
|
||||
- **MCP Server** - Connects Cursor to `https://mcp.algolia.com/1/docsearch/mcp`.
|
||||
- **Rule** - Nudges the agent to use DocSearch for public developer docs instead of guessing from training data.
|
||||
- **Skill** - Documents the one-shot `search_docs` lookup and the `resolve_docset` + `query_docs` flow.
|
||||
|
||||
## Installation
|
||||
|
||||
For local beta testing, install this plugin package from:
|
||||
|
||||
```text
|
||||
mcp/plugins/docsearch/cursor/algolia-docsearch
|
||||
```
|
||||
|
||||
The MCP server config is also available at:
|
||||
|
||||
```text
|
||||
mcp/plugins/docsearch/cursor/algolia-docsearch/mcp.json
|
||||
```
|
||||
|
||||
## Available Tools
|
||||
|
||||
### algolia_docsearch_search_docs
|
||||
|
||||
One-shot lookup, and the right default for most questions. Pass a `library` (product or platform) and a `query` (the question); it resolves the best documentation set and returns ranked content in a single call.
|
||||
|
||||
### algolia_docsearch_resolve_docset
|
||||
|
||||
Step 1 of the manual flow. Finds matching documentation sets and returns candidates with a `docset_id`, title, description, and ranking signals.
|
||||
|
||||
### algolia_docsearch_query_docs
|
||||
|
||||
Step 2 of the manual flow. Queries documentation content for the `docset_id` values returned by `algolia_docsearch_resolve_docset`. Pass several `docsetIds` when a question spans multiple products.
|
||||
|
||||
## Usage Examples
|
||||
|
||||
Ask Cursor questions like:
|
||||
|
||||
- "How do I configure middleware in Next.js?"
|
||||
- "Show me Stripe webhook signature verification from the docs."
|
||||
- "What is the current Algolia InstantSearch React setup?"
|
||||
|
||||
The plugin is for public developer documentation only. It does not expose crawler operations, DocSearch administration, private docs, or Algolia index credentials.
|
||||
|
|
@ -0,0 +1 @@
|
|||
<svg xmlns="http://www.w3.org/2000/svg" viewBox="-10 -10 96 98"><path fill="#5468ff" d="M36.493 77.289H3.074A3.067 3.067 0 0 1 0 74.228h36.523c12.769.062 24.6-6.663 31.038-17.642a35.45 35.45 0 0 0 .174-35.584C61.404 9.962 49.639 3.122 36.87 3.061H0A3.067 3.067 0 0 1 3.074 0H36.87c10.31.01 20.193 4.1 27.473 11.369a38.6 38.6 0 0 1 11.341 27.384c-.06 21.366-17.741 38.536-39.19 38.536z"/><path fill="#5468ff" d="M0 69.045h23.711a11.78 11.78 0 0 0 7.933-3.031H0zM0 60.8h36.374c.734-.987 1.418-1.975 2.053-3.07H0zm0-8.254h41.025a30 30 0 0 0 1.091-3.06H0zm0-8.244h43.306c.149-.987.268-2.034.337-3.061H0zm0-8.244h43.633a30 30 0 0 1-.337-3.071H0zm0-8.254h42.116a36 36 0 0 0-1.09-3.061H0zm0-8.245h38.427a33 33 0 0 1-2.053-3.06H0zM0 8.244v3.071h31.674a19.04 19.04 0 0 0-7.934-3.031z"/></svg>
|
||||
|
After Width: | Height: | Size: 784 B |
8
mcp/plugins/docsearch/cursor/algolia-docsearch/mcp.json
Normal file
8
mcp/plugins/docsearch/cursor/algolia-docsearch/mcp.json
Normal file
|
|
@ -0,0 +1,8 @@
|
|||
{
|
||||
"mcpServers": {
|
||||
"algolia-docsearch": {
|
||||
"type": "http",
|
||||
"url": "https://mcp.algolia.com/1/docsearch/mcp"
|
||||
}
|
||||
}
|
||||
}
|
||||
|
|
@ -0,0 +1,22 @@
|
|||
---
|
||||
description: Use Algolia DocSearch MCP for current public developer documentation instead of guessing from training data.
|
||||
alwaysApply: true
|
||||
---
|
||||
|
||||
When the user asks about public developer documentation, APIs, SDKs, libraries, frameworks, configuration, or code examples, use Algolia DocSearch MCP to fetch current documentation before answering.
|
||||
|
||||
Use Algolia DocSearch when:
|
||||
- The user asks about a public library, framework, SDK, API, or developer tool.
|
||||
- The user needs syntax, configuration, migration, setup, or code examples from docs.
|
||||
- The question may depend on recent documentation changes.
|
||||
- You need source URLs to support an answer.
|
||||
|
||||
Do not use Algolia DocSearch for:
|
||||
- Private company docs, internal repos, or unpublished APIs.
|
||||
- Questions already answered by code or docs the user provided.
|
||||
- Crawler operations, DocSearch administration, or MCP server implementation details.
|
||||
- General programming concepts that do not require external docs.
|
||||
|
||||
For most lookups call `algolia_docsearch_search_docs` with the product as `library` and the question as `query`. Use the `algolia_docsearch_resolve_docset` + `algolia_docsearch_query_docs` flow when a question spans several products or you need to hand-pick documentation sets. Natural-language queries work well.
|
||||
|
||||
See the `algolia-docsearch-mcp` skill for the full tool flow.
|
||||
|
|
@ -0,0 +1,72 @@
|
|||
---
|
||||
name: algolia-docsearch-mcp
|
||||
description: Use this skill when the user asks about public developer documentation, SDKs, APIs, libraries, frameworks, setup, configuration, or code examples. Fetch current docs from Algolia DocSearch MCP and cite source URLs.
|
||||
---
|
||||
|
||||
When the user asks about public developer documentation, use Algolia DocSearch MCP instead of relying only on training data.
|
||||
|
||||
## When To Use
|
||||
|
||||
Use this skill when the user:
|
||||
|
||||
- Asks setup or configuration questions for a public library, framework, SDK, API, or developer tool.
|
||||
- Requests code examples that should match current docs.
|
||||
- Needs API reference details, migration guidance, or command syntax.
|
||||
- Mentions a library or framework whose docs may have changed recently.
|
||||
- Asks for sources or citations from public documentation.
|
||||
|
||||
Do not use this skill for private company docs, internal repositories, unpublished APIs, or questions where the user already supplied the relevant docs.
|
||||
|
||||
## Tools
|
||||
|
||||
DocSearch MCP exposes three tools. You can query in natural language — full sentences and questions work well.
|
||||
|
||||
- `algolia_docsearch_search_docs` — one-shot lookup. Use this by default.
|
||||
- `algolia_docsearch_resolve_docset` — step 1 of the manual flow: find documentation sets.
|
||||
- `algolia_docsearch_query_docs` — step 2 of the manual flow: fetch content for chosen docsets.
|
||||
|
||||
## Default Flow: One-Shot Search
|
||||
|
||||
Call `algolia_docsearch_search_docs` for most lookups.
|
||||
|
||||
Input:
|
||||
|
||||
- `library`: the product, library, SDK, or platform. Use the official name (for example: `Next.js`, `Stripe`, `Algolia InstantSearch`).
|
||||
- `query`: the question, in natural language (for example: `how do middleware matchers work`).
|
||||
|
||||
It resolves the best matching documentation set and returns ranked content in a single call. If `library` is ambiguous, it returns candidate documentation sets instead — pick the right one and pass its `docset_id` to `algolia_docsearch_query_docs`, or retry with the official product name.
|
||||
|
||||
## Manual Flow: Resolve, Then Query
|
||||
|
||||
Use the two-step flow when a question spans several products, or when you want to inspect and hand-pick documentation sets.
|
||||
|
||||
### Step 1: Resolve The Docset
|
||||
|
||||
Call `algolia_docsearch_resolve_docset`.
|
||||
|
||||
Input:
|
||||
|
||||
- `query`: describe the product, library, SDK, or platform. Including the vendor name improves matching.
|
||||
- `topN`: optional. Use the default unless the initial result is ambiguous.
|
||||
|
||||
Each candidate returns a `title`, `docset_id`, `description`, and ranking signals (`trustScore`, `benchmarkScore`, `popularityScore`). Prefer the official vendor's docset over third-party mentions, breaking ties by the higher scores.
|
||||
|
||||
### Step 2: Query The Docs
|
||||
|
||||
Call `algolia_docsearch_query_docs`.
|
||||
|
||||
Input:
|
||||
|
||||
- `query`: what to find in the docs, in natural language.
|
||||
- `docsetIds`: an array of the `docset_id` values you selected. Pass several when a question spans multiple products.
|
||||
|
||||
## Answer With Sources
|
||||
|
||||
Use the returned documentation content to answer the user directly.
|
||||
|
||||
Guidelines:
|
||||
|
||||
- Include source URLs from the tool result when available.
|
||||
- Say when DocSearch does not return a relevant result.
|
||||
- Do not invent source citations.
|
||||
- Keep the answer focused on the user's question.
|
||||
Loading…
Reference in a new issue