AL-Naqawa Skill & MCP: Prayer Tools for AI Agents

Your prayer tools can now work inside an AI conversation

AL-Naqawa Skill & MCP is now available in public beta. Connect a compatible AI agent to calculate local prayer times, convert calendar dates or generate a timetable without copying numbers between websites. The agent can request a structured result from our tools and then help you work with it. The cover is an original illustration of that workflow.

A useful request might be: ‘Find New York City in the United States, confirm its time zone, and create a three-day prayer timetable starting 24 October 2026 using Muslim World League and Standard Asr. Keep the calculation settings in the file.’ The important part is that the assistant calls a calculation tool rather than supplying times from memory.

This launch is for supported agent workflows. It does not mean that every AI app can connect, that an assistant is always correct, or that our planned mobile app and physical prayer clock have launched.

What is the difference between the Skill and the MCP server?

MCP, short for Model Context Protocol, gives compatible AI applications a standard way to discover and call tools. Our remote server exposes named operations with defined inputs. It supplies calculations and data; the AI application handles the conversation. The protocol's own architecture guide explains the client, host and server roles [1].

The AL-Naqawa Skill is a small downloadable package of instructions, capability references and a Python helper. It teaches an agent how to confirm a city, choose settings and use the tools responsibly. It is not a model, a phone app or an automatic installation into your assistant. A client that supports skills and Python execution can use the helper; a client with remote MCP support can connect directly.

You can use the MCP server without installing the Skill. You can also use our regular website without either one. Client support, local execution permissions and any AI subscription charges are determined by your chosen application.

The ten tools available at launch

These are the tools exposed by the live service, rather than a list of proposed features. Their current reference and examples are on our Skill & MCP page [2].

How to connect your assistant

Open the Skill & MCP page and copy https://alnaqawa.com/mcp into your client's remote MCP server or connector settings. Choose Streamable HTTP if the client asks for a transport. The public beta does not require an AL-Naqawa account or API key. The page also provides configuration text for clients that accept an mcpServers entry; other applications have their own setup interface.

After connecting, ask the agent to list the available AL-Naqawa tools. Start with city search, not a guessed city ID. Confirm the country and returned time zone, then state the local date, calculation method and Asr rule before requesting a timetable.

For the Skill route, download the ZIP from the same page, review SKILL.md and follow your client's skill installation instructions. The package includes a Python standard-library helper and a published SHA256 checksum. The helper can save a generated timetable to a file you specify and refuses to overwrite an existing file.

Three practical things to try

Time-zone conversion also helps when clocks change. In a live check, 2026-11-01T05:30:00Z and 2026-11-01T06:30:00Z both mapped to 01:30 in America/New_York, but with UTC offsets of −04:00 and −05:00 respectively. Supplying the original timestamp and retaining the offset distinguishes the repeated hour. A bare ‘1:30 am in New York’ would not be enough information.

• Plan a working week: confirm your city, request a dated timetable with your usual method, then ask the agent to identify the times you want to keep in view. Review the output before arranging meetings.
• Understand a discrepancy: compare MWL and ISNA for one city and one date while keeping the Asr rule and adjustments unchanged. Ask which entries differ, rather than asking the AI to decide which method is religiously correct.
• Prepare a calendar file: request a CSV for a spreadsheet or an ICS for a calendar app. Check the city, dates and method before importing. Importing a file is a separate action in your calendar application.

Keep the calculation context attached

Prayer results depend on the place, local date, time zone and selected conventions. The tools support named presets, Standard or Hanafi Asr, high-latitude settings and minute adjustments. A larger method list does not establish greater religious authority or guarantee a closer match to your mosque. Check which settings your community follows.

The calendar tools support Umm al-Qura and civil calculations with a −2 to +2 day adjustment. These are calculated date labels for planning. Local moon-sighting decisions and community announcements may differ, and Hijri days begin at sunset. The converter is not a local observance announcement.

A calculated prayer start is also different from a mosque's Iqamah or Jumuah schedule. Missing astronomical events are not an invitation for the agent to invent a time. Ask it to retain any warnings and explain unavailable results.

What the beta does—and what still needs your review

The current service is read-only. It does not send emails, create appointments in your account, set phone alarms or control a prayer clock. An ICS file contains prayer start markers, without a built-in alarm or a claimed prayer duration. Your calendar application may apply its own defaults, so inspect the imported events.

We checked tool discovery, the live method list, a dated New York City calculation, a method comparison, calendar conversion, the repeated New York daylight-saving hour, and three-day CSV and ICS generation for this article. These checks establish that the requested workflows returned results; they are not an independent astronomical accuracy study or a guarantee that every AI client handles them identically.

City search covers the cities currently available in the service; the complete source catalog is still being imported. The beta has bounded requests and no unlimited-use or availability promise. Some measured calls have exceeded the usual free-tier CPU budget, so a request may fail or need to be made smaller. Do not build an unattended critical reminder system around a beta response.

The MCP service does not create a personal profile or a new stored conversation history. Requests still pass through Cloudflare, and your AI client applies its own data and logging policies. Avoid putting private personal information into a calculation request.

Start with one clear, checkable request

Connect your assistant, confirm the right city and try a single dated calculation. Check the result against the settings on our website and the timetable you normally follow. Once that is clear, combine tools for a week, a trip or a file you can keep.

The aim is practical: make the same prayer tools useful in the place where you are already planning your day, with enough context to understand the answer.