Discover MCPs & agents
Loading MCPs and agents…
Loading MCPs and agents…
Mobile UI/UX design research: real app screens, user flows and UI patterns with cited evidence.
From the repo.
A remote, read-only Model Context Protocol server for mobile UI/UX design research. Built for product designers and developers, it gives MCP-compatible assistants (Claude, Cursor, Codex, Windsurf and others) access to Flownato's library of real mobile app screens, user flows and UI patterns, with every answer tied to published screen evidence.
https://agent.flownato.com/mcp (Streamable HTTP)Flownato records real mobile apps such as Swiggy, Zepto, BookMyShow, redBus and Nykaa, then publishes reviewed screens, complete journeys and a UI pattern taxonomy. The MCP server lets an assistant search and compare that evidence instead of guessing what an interface looks like.
Typical questions it can answer with citations:
| Tool | Purpose |
|---|---|
whoami | Reports the authenticated membership, what that tier can see, and the client. |
search_ui_library | Finds concrete journey examples for a research question. |
search_ui_screens | Screen-level search with exact interface text matching. |
analyze_ui_pattern | Measures one screen pattern, UI element, flow pattern or flow action across an optional app or category scope. |
explore_ui_patterns | Discovers recurring patterns when you do not yet know which one to analyze. |
compare_ui_flows | Compares the same product moment across two or three named apps. |
inspect_app_library | Summarizes one app's published coverage, patterns, elements and journeys. |
check_ui_presence | Scoped yes/no presence checks with evidence. |
inspect_library_coverage | Paginated inventory of what is published. |
inspect_ui_resources | Exact release-bound follow-up retrieval for returned resourceRef objects. |
get_screen_image | Returns a published screen image inline after another tool supplies its asset ID. |
All tools are read-only. Results cite the exact journey and step identifiers they were built from, and pattern counts are descriptive, never presented as best practice.
The plugin installs the MCP server together with the flownato-ui-research skill, which tells the assistant when
to look up real app flows and how to cite them:
/plugin marketplace add flownato/flownato-mcp
/plugin install flownato@flownato
The same repository carries plugin manifests for Cursor (.cursor-plugin/) and Codex (.codex-plugin/). The skill
lives in skills/flownato-ui-research/SKILL.md.
In clients that accept a remote mcpServers configuration, use:
{
"mcpServers": {
"flownato": {
"type": "http",
"url": "https://agent.flownato.com/mcp"
}
}
}
Claude Code:
claude mcp add --transport http flownato https://agent.flownato.com/mcp
On first use the client opens https://api.flownato.com/oauth/authorize in your browser. Sign in with your Flownato account and approve read-only access. Tokens are short lived and refresh silently until you revoke the grant.
Discovery documents:
https://agent.flownato.com/.well-known/oauth-protected-resource/mcphttps://api.flownato.com/.well-known/oauth-authorization-server| Plan | What the MCP returns |
|---|---|
| Not signed in | The endpoint answers 401 with OAuth discovery metadata, so your client opens Flownato sign-in. Creating an account is free. |
| Free account | All eleven tools. Journeys marked free on flownato.com (up to three per app) return every screen. Other journeys return their first 3 screens; the rest come back with locked: true, no asset ID and no image URL. Every result carries an access block with the locked counts and an upgrade link, and get_screen_image answers PREMIUM_REQUIRED for a locked screen. |
| Premium | All eleven tools with the complete library. |
Both tiers are metered at a default safety limit of 500 tool calls per UTC day.
Prompts, arguments and results are not stored. Screenshot URLs are signed and short lived, and the image endpoint rechecks membership when the image is fetched.
The server reads only the published product catalog: reviewed, immutable releases with privacy redaction applied. It has no access to raw crawl data, model output or other users' accounts. It answers UI and product-design research questions and declines unrelated requests.
This repository documents the hosted service and holds its plugin manifests and skill. The server itself is part of Flownato's private product codebase (Python, FastAPI) and is not open source.
lhm.plugin.json declares the existing LobeHub listing's category,
remote endpoint, and all eleven MCP tool schemas. Its category is developer
(displayed as Developer Skills on LobeHub), reflecting its use for UI/UX design
and product development.
The endpoint requires OAuth before protocol initialization and tool discovery.
An unauthenticated request returning 401 is expected; use the protected-resource
metadata above to start sign-in. Directory crawlers cannot discover the tool list
from an unauthenticated connection alone.
The server currently exposes tools, with no MCP prompt templates or MCP resources.
Returned resourceRef objects are arguments to inspect_ui_resources; they are
not entries in the MCP resources/list capability. Keep marketplace declarations
consistent with the server's actual capabilities.
The public badge above supports LobeHub's README claim-status check. Prefer this public verification route when available. Updating this public listing does not justify granting a directory read/write access to private GitHub repositories.
Already verified maintainers can update the existing listing with the official
@lobehub/market-cli (Node.js 22 or newer):
# Sign in to LobeHub, then check existing GitHub verification:
npx -y @lobehub/market-cli login
npx -y @lobehub/market-cli github status
# Requires ownership verification; run after reviewing lhm.plugin.json:
npx -y @lobehub/market-cli plugin update --dir "$PWD"
Review requested permissions before connecting any GitHub account. If a claim or
update requires broad repo or workflow access, leave that connection disabled
and request a narrower verification method from LobeHub.
When tool signatures change, refresh tools from the authenticated server's
tools/list response before publishing. Keep prompts and resources empty
unless the server implements them. The listing version remains 1.0.0 for this
metadata correction; it does not represent a new server release. Verify the
public listing
after each update.
Use the endpoint URL below in your config. No API key — you connect directly.
Tool
OS
Config file: ~/.cursor/mcp.json
{
"mcpServers": {
"mcp-server": {
"url": "https://agent.flownato.com/mcp"
}
}
}Paste into mcpServers in the config file. Restart Cursor after saving.
If this MCP is also published on mcpchannel.ai, you can subscribe from Browse and use the gateway config there instead.