Create an AI tool you can text with.

What the AI will do

• Finds an AI activity you would like to reach while away from your desk, using your actual projects and recurring requests.
• Inspects your existing tools, researches the messaging connection that fits, and builds and tests the exchange from your phone through to the result.

What you can provide

• An existing tool or recurring need, if you have one in mind. Otherwise, access to relevant work and a little about the messaging app you use is enough.

Where you'll take part

• Choose the useful activity and messaging route, complete any account steps only you can perform, and send a few test messages.

What you'll get

A working contact or bot you can text to trigger AI work on the run and receive the result, with a simple way to pause it. If a required connection remains blocked, the AI identifies the unfinished part.

Activity context

You are heading into a meeting and want the latest on a project. Or an idea occurs to you between errands and you want your AI to put it where it belongs. Send a quick message, let the tool do the work, and get on with your day. This activity adds that easy point of access to something useful—possibly a tool you already have.

For the user

Start with the prompt below. You do not need to choose Twilio, Telegram, WhatsApp, or any other service in advance. The AI investigates your setup, proposes a useful exchange, and does the building. You help with the choices and account access it cannot handle, then try the tool from the phone you will actually use.

Guidance for the AI

These are general instructions, not a description of this user's particular AI system, tools, access, or circumstances. Adapt the method to the user's goal and the capabilities you can actually use, while preserving the activity's essential sequence, review points, and intended result.

Start with the activity and the user's context

Read the complete activity and guide the user through it. Use the activity prompt yourself; the user does not need to paste it again. Keep the responsibilities of the user, the AI, and any colleague clear.

Begin with what the user has already supplied. Establish the intended result and resolve missing context through focused questions when needed. Work within the activity's stated scope, permissions, required confirmations, and stopping points, including any deliberate order for introducing information. Make progress with the available information and reasonable, clearly identified assumptions where the activity allows them.

Check the capabilities needed for this work

Inspect relevant available tools, connectors, and data. Use small, relevant, non-destructive checks to establish whether necessary access or a capability actually works. Naming a product does not establish access to it. Let this activity determine what needs checking; do not impose a universal preliminary questionnaire or a full system audit.

When useful access is missing:

• Explain what you cannot reach and how that material would help.
• Investigate practical setup, connection, export, or other access alternatives before asking the user to gather material or troubleshoot.
• If progress depends on access, request only the specific user action or permission needed to enable it or provide the material another way. Continue useful independent work while awaiting the response.
• Offer to walk the user through an export or upload. Welcome files, links, copied text, rough notes, or an interview as supported by your system and the activity, and use the best available route. If the missing access is optional, continue with what is available.

Ask focused questions and respect the answers

Ask questions only when necessary to complete the work well or satisfy the activity's required review points. Build on answers already supplied. Ask one question at a time, or a small group of closely related questions. Wait on work that depends on the answer, and continue useful work that does not. Explain unfamiliar language with examples where helpful.

Confirm answers or decisions you supply on the user's behalf before treating them as the user's final answers or decisions. This does not require separate approval of every sourced finding; follow the activity's specific review requirements.

Keep evidence and independent contributions distinct

Distinguish evidence, inference, and missing information. Check your work against the source material.

Use independent agents when your system actually supports delegation and independent investigation or review would improve the work:

• Give each agent a bounded brief and the context appropriate to its role.
• Keep discovery independent until synthesis where the activity requires it. Preserve any requirement for the user or colleagues to contribute before seeing other explanations.
• Follow the activity's stated fallback and any required confirmation when independent agents or fresh contexts are unavailable. Preserve reviewer isolation where required; describe shared-context or sequential passes accurately.

Produce, check, and revise the result

Do the substantive investigation and production. Carry the activity through to the most complete, usable result within its scope, rather than ending with a plan or handing the main work back to the user. When missing information does not prevent useful work, proceed with clearly identified assumptions where the activity permits them.

Check the result against the activity's completion criteria. At each substantive delivery, state what is complete and any material unfinished work, assumptions, helpful missing information, and remaining uncertainty. Explain how these affect what the user can use, and help the user revise the result.

Preserve the activity's stopping point. Do not turn a rehearsal, investigation, or test-design activity into execution that the user has not authorized.

Your AI prompt

Build a way to reach AI while I am on the run

Help me create a tool I can use by sending a quick text, without opening my AI app. Find something useful in my actual work, connect the messaging channel to the AI activity, and make the complete exchange work. I might already have the underlying tool; the valuable addition is being able to reach it in a moment between meetings or while away from my desk.

Start with what you can inspect. Do not require me to invent a texting use case or choose a vendor before you understand my setup. Ask focused questions one at a time when a real gap affects the next step.

Find the activity and the connections I already have

Try the relevant available connections and inspect their actual capabilities. Search recent email, chats and other accessible messages, calendar entries, project notes, and my existing AI conversations or tool documentation. Look for quick requests I make repeatedly, information I need before meetings, ideas I capture on the move, and useful tools I already use. Follow the relevant project and person names into their records. Keep this investigation bounded to identifying a promising activity and its required context; do not sweep my whole organization unnecessarily.

For example, I might text “What changed since my last conversation with Dana?”, “Save this idea under the launch project,” or “Find the latest decision about the London visit.” These are illustrations. Propose a small number of possibilities grounded in my records, explain what each would do and what evidence made it promising, and let me choose or supply my own. If I have already named the activity, work from that.

Inspect my existing messaging and automation arrangements, accessible applications, subscriptions, deployed tools, and relevant setup documentation. Establish which messaging app I actually use, which account owns the tool, where it runs, what the AI can read or change, and whether it already offers a supported message trigger. A connected search tool does not necessarily provide a programmable incoming-message event or permission to send replies. Test those capabilities separately.

If a source is unavailable, first investigate the permitted connection or export options, then request the specific missing access or material and explain what it enables. Screenshots or descriptions of settings can help establish the setup; never ask me to paste a password or API secret into our conversation.

Research a route that fits this setup

Use current official documentation for the actual versions, country, account type, and organization involved. Prefer extending a working supported integration when it fits. Research alternatives only as far as needed to find a usable route. A webhook is simply a way for the messaging service to notify our running tool when a message arrives; it is one possible connection, not the whole tool.

Relevant starting points, checked during authoring on September 13, 2026, include:

• Twilio SMS quickstart and current SMS trial restrictions. Check the actual account and number: trial interfaces differ, and the newer trial can restrict custom text. Do not promise arbitrary AI replies on a trial that does not support them.
• Telegram's bot tutorial and Bot API. A bot can receive messages through either a webhook or long polling; choose one. Establish the bot identity, the user's chat, the token setup, and the process that will handle messages.
• WhatsApp through Twilio. This uses the business platform, with account and sender setup; having the consumer app alone does not supply an API. Check the current rules for freeform replies, templates, and any delayed result before choosing it.

These are research leads, not a requirement to open three new accounts. Read the live documentation and inspect existing controls before recommending setup. If the organization has a better supported channel, investigate it. Preserve my preferred quick-message experience; explain any required change of messaging app and obtain my choice.

Show the recommended exchange in plain language: my message, where it arrives, what AI activity runs, which records it uses, and where the result returns. Establish the runtime and model access, expected ongoing costs, any account verification, and the few setup actions only I or an administrator can perform. Check whether the chosen AI subscription actually includes the execution interface needed; do not infer API access from access to a chat app. Get agreement on the concrete route and any new paid service before committing to it.

Build the complete exchange

Implement and configure everything you can within the agreed scope. Ask me only for the necessary authentication, account ownership, or approval steps, with instructions matched to the actual screen or service. Resume the build after each such step. A diagram, sample webhook, or request to IT is an intermediate result.

Connect incoming messages to the selected activity and return its real result to the same user in the agreed channel. Verify that the deployed process can reach the same permitted sources used in our demonstration; access in this chat does not automatically carry over. Reuse the existing tool's behavior where possible. For a new activity, define its inputs, source searches, output, and permitted actions explicitly.

Make short everyday language work. If a request is ambiguous, have the tool ask a concise clarification in the same conversation and retain enough state to use the answer. If processing takes time, acknowledge receipt, run the work outside the channel's response deadline, and send the result when ready using a supported delivery route. Check any channel timing or template rule affecting that later message. Provide a readable short answer with an accessible artifact link when the result is too long for text.

Authenticate the incoming integration and restrict use to the agreed sender or users. Use the provider's documented verification method, such as Twilio request validation. Store credentials through the host's secret mechanism. Keep untrusted text from changing the tool's authority. Record message and job identifiers so a repeated delivery does not repeat a write or send two results. Give the user a clear failure message when a task fails; never report “saved” or “sent” without checking the underlying action.

Agree what requests can execute immediately and which require a confirmation in the message exchange. Implement that distinction. A tool that can save my notes does not thereby gain authority to email colleagues or change company records.

Use a bounded integration specialist or a fresh reviewer where available: give them the chosen route, relevant official documentation, exact configuration or code, and acceptance cases. Have them find missing connections and test assumptions. Keep one coordinator responsible for deployment and actions. If separate agents are unavailable, do distinct implementation and verification passes and state that they share context.

Test from my phone and keep it running

Prepare the tests yourself and perform those your tools support. Have me send a small number of actual messages from the phone and account I will use. Check the received event, actual AI work, resulting artifact or change where applicable, and the response arriving in that same message conversation. An API accepting a send is not evidence that I received it.

Include a normal request, a follow-up clarification, a repeated event, an unavailable source or failed job, and an unauthorized sender where test facilities permit. Test a consequential request against its agreed confirmation rule. Verify long-running results through the chosen channel as well as short answers. Repair failures and rerun the affected exchange.

Deploy on a runtime that continues when this chat ends and, where intended, when my laptop is asleep. A temporary development tunnel or an open terminal is not a durable on-the-run tool. Verify startup, stored state, and recovery after a restart; record the limits of what was actually tested.

Finish with the contact or bot I can save, a few useful example texts based on the selected activity, what it can do, and a simple way to pause it or check a failed request. Show the evidence of a successful real round trip and identify any outstanding dependency plainly. If deployment, account access, or delivery remains blocked, preserve the working pieces and exact next action, but call the integration unfinished. Complete this activity when I can quickly message the tool, have it perform the agreed AI work, and receive its result without opening the AI app.

Done when

You can send a quick message from your usual phone, have the tool perform the agreed AI activity, and receive its actual result in the chosen conversation without opening the AI app. The service continues beyond the setup chat, and you know how to pause it. A code sample or unfinished messaging connection does not yet complete the move.

These instructions were written in September 2026 using the frontier models available at the time. You may now have better tools or models, or restrictions that make some tasks harder. Your AI should adapt to the tools and access you actually have and find the best available way to help. Everything here is a starting suggestion for you and your AI, not a proven best practice for every setup. Feel free to rewrite any of it to better meet your needs.