mobile-docs-mcp
A version-aware documentation MCP server for mobile development that verifies API availability against target platform versions.
README
mobile-docs-mcp
A version-aware documentation MCP server for mobile development — the context7 idea, but specialized for the thing generic doc proxies get wrong: which version of an API you can actually use.
Install:
uvx mobile-docs-mcp(nothing to clone) — orpip install mobile-docs-mcp. Published on PyPI. Jump to Editor & IDE setup.
It indexes Android (Jetpack / androidx / Compose) and Apple (SwiftUI / UIKit)
documentation with since / deprecated / removed metadata on every symbol,
so an agent can verify an API exists on the version the project targets before
writing code against it.
Why it's different from a doc proxy
Most "Apple docs" / "Android docs" MCP servers live-fetch and parse the current
doc site. They answer "what is scrollTargetBehavior?" but not "can I use it on
iOS 16?" — and that second question is where coding agents hallucinate. Every
result here is filtered and annotated against a target version.
Tools
| Tool | Purpose |
|---|---|
search_mobile_docs |
Hybrid (BM25 + optional vector) retrieval, filtered by platform + target version |
get_api_reference |
Full symbol card: signature, availability, deprecation/migration pointer |
verify_api_exists |
Anti-hallucination check — real AND usable on a given version? |
list_api_versions |
since / deprecated / removed timeline + migration lineage |
get_release_notes |
What changed in a library/version |
Example verdict:
verify_api_exists("scrollTargetBehavior", "16.0", "apple")
→ ❌ NOT AVAILABLE — introduced in 17.0, newer than 16.0. Will not compile there.
Run
# Recommended: fetch + run on demand, no install step
uvx mobile-docs-mcp
# Or install into the current environment
pip install mobile-docs-mcp
mobile-docs-mcp # stdio transport
# From source (for development)
git clone https://github.com/asif786ka/mobile-docs-mcp
cd mobile-docs-mcp && pip install -e .
python -m mobile_docs_mcp.server
Ships with a seed corpus, so it works immediately with no crawl.
Editor & IDE setup (iOS + Android)
The server speaks MCP over stdio, so any MCP-capable editor can use it. All the tools work for both platforms — you'll lean on the SwiftUI entries in an iOS project and the Compose/androidx entries in an Android one, but nothing is editor-specific. Two ways to launch it:
- Recommended (from PyPI):
command: "uvx",args: ["mobile-docs-mcp"]— no clone, nocwd. This is what the examples below use. - From source:
command: "python",args: ["-m", "mobile_docs_mcp.server"], withcwdset to the repo path.
Set MOBILE_DOCS_DATA in env to layer in any corpora you've crawled with the
ingesters (see "Growing the index").
Claude Code (CLI)
# published package, shared with the repo via .mcp.json (project scope)
claude mcp add mobile-docs-mcp --scope project -- uvx mobile-docs-mcp
# from source instead: claude mcp add mobile-docs-mcp -- python -m mobile_docs_mcp.server
Verify with claude mcp list, or /mcp inside a session.
Claude Desktop
Edit claude_desktop_config.json (Settings → Developer → Edit Config), then
restart the app:
{
"mcpServers": {
"mobile-docs-mcp": {
"command": "uvx",
"args": ["mobile-docs-mcp"]
}
}
}
(Layer in crawled corpora with "env": { "MOBILE_DOCS_DATA": "/abs/path/data/android,/abs/path/data/apple" }.)
Cursor
Create .cursor/mcp.json in the project root (or ~/.cursor/mcp.json for all
projects) — same mcpServers shape as above:
{
"mcpServers": {
"mobile-docs-mcp": {
"command": "uvx",
"args": ["mobile-docs-mcp"]
}
}
}
Then enable it under Settings → MCP.
Windsurf
Edit ~/.codeium/windsurf/mcp_config.json (Cascade panel → MCP icon →
Configure). Same mcpServers shape; Windsurf supports ${env:VAR} interpolation
so you can keep paths/secrets out of the file:
{
"mcpServers": {
"mobile-docs-mcp": {
"command": "uvx",
"args": ["mobile-docs-mcp"]
}
}
}
Android Studio
Gemini in Android Studio is an MCP client. Go to Settings → Tools → AI → MCP
Servers, tick Enable MCP Servers, and paste the same mcpServers JSON
block shown for Claude Desktop. Confirm with the "Successfully connected"
notification, then type /mcp in the chat to see the tools. Great for checking
androidx/Compose availability against your module's compileSdk / library
versions without leaving the IDE.
VS Code
VS Code's key is servers (not mcpServers) and wants an explicit type.
Create .vscode/mcp.json (commit it to share with the team):
{
"servers": {
"mobile-docs-mcp": {
"type": "stdio",
"command": "uvx",
"args": ["mobile-docs-mcp"]
}
}
}
Open the file and click Start, or run MCP: List Servers from the Command Palette. Copilot Chat's Agent mode will then surface the tools.
Xcode
Xcode 26.3+ is itself an MCP server (it exposes build / test / preview /
Apple-doc-search tools) rather than an MCP client — so you don't register
mobile-docs-mcp inside Xcode. Instead, run an agent that consumes both: point
Claude Code (or Cursor / Codex) at mobile-docs-mcp using the steps above while
it's also connected to Xcode's server. The agent then gets Xcode's build/preview
tools and this server's verify_api_exists version checks in one session —
e.g. it can confirm scrollTargetBehavior needs iOS 17 before writing it against
your iOS 16 deployment target, then build the project to check.
Architecture
ingest/ (batch, offline) server.py (runtime)
android.py developer.android.com FastMCP
apple.py Apple JSON doc API │
│ ▼
▼ store.py — hybrid, version-aware
symbols.json + chunks.json 1. BM25 lexical recall
(version-tagged) 2. vector recall (optional)
3. RRF fusion
4. VERSION FILTER ← the differentiator
5. cross-encoder rerank (optional)
Retrieval is provider-agnostic (embeddings.py), same shape as a multi-provider
model gateway: BM25 works fully offline; vector recall and rerank switch on via
env var with no code change and degrade gracefully if a provider is missing.
MOBILE_DOCS_EMBEDDINGS=local # or openai ; default none (BM25-only)
MOBILE_DOCS_RERANK=local # default none
MOBILE_DOCS_DATA=data/android,data/apple # layer in crawled corpora
Growing the index
python -m mobile_docs_mcp.ingest.android --out data/android \
--refs androidx.compose.foundation.lazy.LazyColumn \
--release-notes compose-foundation navigation
python -m mobile_docs_mcp.ingest.apple --out data/apple \
--paths swiftui/navigationstack swiftui/view/scrolltargetbehavior
Then point the server at the output with MOBILE_DOCS_DATA.
Status
MVP. The protocol layer, version logic, hybrid store, and tool surface are production-shaped; the ingestion coverage is a seed. The real work — and the moat — is breadth and freshness of the version-tagged index.
MIT.
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.
Neon Database
MCP server for interacting with Neon Management API and databases
E2B
Using MCP to run code via e2b.
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.