Works with the AI Agents you already use
Add MCP Server: Ask Claude About Your App
A tool that re-derives a figure, leaks another tenant's row or accepts a write is worse than no tool.
Free
Add MCP Server: Ask Claude About Your App
Example session with this skill installed
MCP-ify my Laravel booking app so the owner can ask Claude about yesterday's takings and what's low in stock. Multi-tenant, one database per tenant.
- Read your context and instructions
- Compiled the add mcp server
Summary
Inventory done. Here is the proposed tool list for your approval. I have not written code yet, because it is cheap to change now and costly after clients cache the schema.
TOOL LIST (read-only, all readOnlyHint: true)
Tool Question Source query Permission
list_branches Which branches can I see? BranchScope::forUser() any
daily_takings How did we do yesterday? Reports/DailySalesReport::run reports.view
low_stock What is low? Reports/StockLevels::lowItems inventory.view
top_services What sold best this week? Reports/ServiceMix::run reports.view
Findings
Scope: tenant from host, then BranchScope::forUser() per request
Auth: sessions only, no OAuth provider -> API keys first, then OAuth 2.1
Time: tenant time zone, never server time
SDK: official PHP MCP SDK, name and version to be checked against current docs
Mount: /mcp inside the existing app, stateless Streamable HTTP
Server instructions: "Read-only figures from the booking app. Call list_branches first. Dates are local dates. Quote every figure; do not estimate or invent figures."
Next steps
- Confirm or edit the four tools.
- I will then build tools, API keys and tests, and add OAuth for claude.ai and ChatGPT.
- claude.ai and ChatGPT need public HTTPS, so I will mark them unverified until tested on a deployed URL.
Connects securely to your tools. The creator never sees your data.
About this skill
Make the tool trustworthy before making it reachable
An assistant quotes whatever a tool returns. A tool that re-derives a figure, leaks another tenant's row or accepts a write is worse than no tool. Add MCP Server builds in a fixed order: read-only first, your app's own queries, your app's own scope rules, then auth, then the clients. The app itself calls no LLM.
How it works
- Inventory before code. Finds the 5 to 10 questions the owner asks, the existing query behind each, the scope model, the auth model and the local-day rule. You confirm the tool list before anything is built.
- Design the tools. Read-only annotations, no re-derived figures, scope checked on every call, server instructions that forbid invention, plain JSON output with a
note, and an audit row per call. - Pick the SDK. The official SDK for your language, over stateless Streamable HTTP, mounted at
/mcpin the existing app. - Authentication. Scoped API keys first, shown once and stored hashed. Then OAuth 2.1 with PKCE, resource binding and refresh-token reuse detection. An existing OAuth provider is reused if there is one.
- Prove it. Tests that tool figures equal the report's, out-of-scope ids look like missing ones, nothing is written, and bad tokens get a 401 challenge. Then the Inspector and a real question in Claude Code.
- Hand over. URL, client setup, what data leaves the app, how to revoke, what was not verified.
Why buyers like it
- Read-only by test, not by promise.
- It tells you what it could not verify.
Need help with it?
This skill is free. It is part of ShipKit, built by Parvez. If you want it set up in your project, adapted to your team's workflow, or a custom skill built for you, get in touch at parvez.cloud.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.
Public HTTPS is required for web-based client integration. Read-only tools only. No support for local stdio tools or write operations. SDK versions for 2026+ require specific verification.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
1 use
Times used to date
Free to use
No account required to browse
Trust & safety
Security scan
Security scan passed 8 days ago
- Free with an Agensi account
- Software
- Ify MCP@modelcontextprotocol/inspector
Node.jsTypes MCP
Python