> For the complete documentation index, see [llms.txt](https://learn.doubletick.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://learn.doubletick.io/doubletick-ai/ai-agents-and-assistance/create-an-ai-agent-for-whatsapp-and-instagram-chats.md).

# Create an AI agent for WhatsApp & Instagram chats

Learn how to create an AI agent that handles your WhatsApp and Instagram chats, from writing its prompt and adding knowledge to connecting tools, testing and publishing it.

Your customers expect a reply the moment they message you. An AI agent for chats replies instantly, at any hour, in 30+ languages, without adding to your team’s workload. It works on **WhatsApp and Instagram** chats. It answers from your own business information, sends rich messages like images, buttons and catalogues, and takes actions such as booking a demo or handing the chat to the right person. You decide what it knows, what it can do and when it must hand over.

This guide walks you through setting one up, step by step. No coding is needed.

#### What's covered

* [Before you begin](#before-you-begin)
* [How to create your AI agent](#step-1-how-to-create-your-ai-agent)
* [How to write a prompt that gets results](#step-2-how-to-write-a-prompt-that-gets-results)
* [How to set up the chat settings](#step-3-how-to-set-up-the-chat-settings)
* [How to decide what happens when the agent isn’t sure](#step-4-how-to-decide-what-happens-when-the-agent-isnt-sure)
* [Give your agent the right knowledge](#step-5-give-your-agent-the-right-knowledge)
* [Let your agent take actions with tools](#step-6-let-your-agent-take-actions-with-tools)
* [How to connect MCP servers](#step-7-how-to-connect-mcp-servers)
* [How to test your agent before going live](#step-8-how-to-test-your-agent-before-going-live)
* [How to publish your agent](#step-9-how-to-publish-your-agent)
* [Use your agent on Instagram](#use-your-agent-on-instagram)
* [How does your agent handle voice notes](#how-does-your-agent-handle-voice-notes)
* [How does your agent handle WhatsApp Flow replies](#how-does-your-agent-handle-whatsapp-flow-replies)
* [Keep improving your agent](#keep-improving-your-agent)

#### Before you begin

* You need the **AI agents and assistance add-on** on the Pro or Enterprise plan.
* Each message the agent sends uses **1 AI credit**. Each voice note it transcribes uses **0.5 AI credits**. For the full rate card, see [Introduction to AI agents](https://learn.doubletick.io/doubletick-ai/ai-agents-and-assistance/introduction-to-ai-agents.md).
* Keep your business information handy: FAQs, your website link, policies and product documents.

#### Step 1: How to create your AI agent

<figure><img src="https://2303112206-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F53n17VnOICC1LtDqlENV%2Fuploads%2FkYUIxYV2o28C1ycf7Ebk%2F01%20-%20Create%20your%20AI%20agent.gif?alt=media&#x26;token=b3697792-505c-41b2-a2fa-6598fa1db11c" alt=""><figcaption></figcaption></figure>

* **Click** **Automate** in the left menu, then **click** **AI agents**.
* **Click** **Create new agent**.
* **Enter** a name your team will recognise, such as “Support” or “Sales”. Customers never see this name.
* **Describe** the agent’s job in one or two lines. This becomes the start of its prompt. You can also click **Skip** and write it later.
* **Select** any shared knowledge bases the agent should answer from. You can also add knowledge later.
* **Click** **Create agent**.

Your agent opens in its editor. The left menu has three groups:

* **Chat config:** **Prompt**, **Setup** and **MCP servers** for chats.
* **Call config:** **Prompt**, **Setup** and **Post call metrics** for calls.
* **General:** **Knowledge base**, **Tools**, **Unresolved intents** and **Versions**, shared by chats and calls.

{% hint style="info" %}
**Note:** Your changes are saved as a draft, and the header shows how many unpublished changes you have. Customers keep talking to the live version of the agent until you click **Publish** (Step 9).
{% endhint %}

#### Step 2: How to write a prompt that gets results

<figure><img src="https://2303112206-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F53n17VnOICC1LtDqlENV%2Fuploads%2FOGLyEhNpKUttW7OKh0uu%2F02%20-%20Write%20a%20prompt.gif?alt=media&#x26;token=7307f633-c0bf-46c5-8deb-e7fcbdd3c615" alt=""><figcaption></figcaption></figure>

The prompt is the most important part of your agent. It tells the agent who it is, what it should achieve and where it must stop. The agent reads it before every chat. A vague prompt gives vague answers. A **clear goal** with **clear rules** gives answers your customers can trust.

* **Click** **Prompt** under **Chat config**.
* **Write** your instructions. Cover these four things:

  **Who the agent is:** its name, role and the business it represents.

  **What it should do:** answer product questions, share order updates, guide customers, and so on.

  **How it should talk:** tone, language, and words to use or avoid.

  **Rules it must follow:** never guess an answer, always hand over complaints, and so on.
* Write it like a document. Use **Formatted** to see headings, lists and tables, or **Raw** to write in Markdown. Click the full-screen icon if you need more room.
* **Type** **/** to turn a line into a heading, a list, a checklist or a quote, or to insert a **Table**, a **Code block**, a **Divider** or a **Tool or knowledge** base by name. You can also click **+ Insert**.

**Sample System Prompt:**

{% code overflow="wrap" expandable="true" %}

```
Role and Identity
You are [Agent Name], the official virtual assistant for [Company Name]. You assist customers with questions about [Company Name]'s products and services. You are available 24 hours a day, 7 days a week.
Always introduce yourself by name when starting a new conversation.
Objective
Your primary goal is to resolve customer queries accurately and efficiently. If you do not have enough information to answer confidently, do not guess. Instead, let the customer know you are connecting them with a team member who can help.
Tone and Language
Always be polite, professional, and easy to understand. Use simple, clear language. Keep responses concise.
If a customer writes in a language other than English, respond in the same language.
If a customer is frustrated or upset, acknowledge their concern before responding. Never argue with a customer.
What You Can Help With
Answer questions related to [Company Name]'s products, services, pricing, processes, and policies — but only if the information is available in your knowledge base. Guide customers to the right team member when a query is beyond your scope.
What You Must Not Do
Do not make up information. Do not make promises about outcomes, timelines, or refunds unless that information is explicitly in your FAQs. Do not discuss competitors. Do not share internal company information. Do not engage with unrelated questions.
Escalation Rules
If a customer asks to speak to a human agent, use the Talk To Human tool immediately without asking further questions. If a customer has a complaint requiring team action, use the Talk To Human tool and inform them a team member will follow up shortly. If it is outside business hours, let the customer know the team will respond as soon as possible.
Closing a Conversation
Once you have answered a customer's query, check if they need anything else. Use a friendly but brief closing line. Do not end abruptly.
```

{% endcode %}

{% hint style="info" %}
Your prompt can be up to 10,000 characters on the Pro and Enterprise plans.
{% endhint %}

#### Step 3: How to set up the chat settings

<figure><img src="https://2303112206-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F53n17VnOICC1LtDqlENV%2Fuploads%2FJNuwnNfNjXTguec8wtid%2F03%20-%20Choose%20the%20model%20and%20settings.gif?alt=media&#x26;token=e5c739c5-91fa-4568-a84f-f9dd85cec4d8" alt=""><figcaption></figcaption></figure>

**Click** **Setup** under **Chat config**. Settings are grouped into sections, such as Conversation, Agent, Web search, Fallback, Customer memory, Custom fields and Customer outcomes. To find a setting quickly, **type** its name in **Search chat settings** at the top and **click** the result to jump to it.

<table data-search="false"><thead><tr><th width="206.00006103515625">Setting</th><th>What it does</th></tr></thead><tbody><tr><td>Model</td><td>The AI model the agent uses. Pick one based on how complex your chats are.</td></tr><tr><td>Context messages</td><td>How many earlier messages in the chat the agent remembers. The default is 5.</td></tr><tr><td>Max reasoning steps</td><td>How many tool calls the agent can make to answer one message.</td></tr><tr><td>Include all tools</td><td>Gives the agent every tool on each reply.</td></tr><tr><td>Reasoning effort</td><td>How deeply the model thinks before it replies. Leave it on Default to let the model decide.</td></tr><tr><td>Show reasoning summary in test mode</td><td>Shows why the agent replied the way it did while you test. This uses slightly more credits.</td></tr><tr><td>Enable web search</td><td>Lets the agent look up live information on the internet.</td></tr><tr><td>Default intent behaviour</td><td>What happens when the agent isn’t confident. See the next section.</td></tr><tr><td>Enable customer memory</td><td>Lets the agent remember each customer across chats. Choose a template, such as Generic customer profile, Support bot, Real estate bot or Shopping bot, to decide what it remembers. It never overwrites your CRM data.</td></tr><tr><td>Custom fields</td><td>Shares the custom fields you select with the agent as facts about the customer.</td></tr><tr><td>Customer outcomes</td><td>The outcomes each AI call can end in, such as “Callback later”. When a call ends, the conversation is matched to one outcome and your bot flow continues from there.</td></tr></tbody></table>

{% hint style="info" %}
**Note:** Some settings, such as Model, Max reasoning steps and Reasoning effort, may not show on every account.
{% endhint %}

#### Step 4: How to decide what happens when the agent isn’t sure

<figure><img src="https://2303112206-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F53n17VnOICC1LtDqlENV%2Fuploads%2FZBcW0motSAlpLzNNZKj3%2F04%20-%20Set%20up%20the%20fallback%20flow.gif?alt=media&#x26;token=1ae3852c-9e27-4f1d-aad3-ac56026771cf" alt=""><figcaption></figcaption></figure>

When the agent can’t find a confident answer, it follows a fallback flow that you build. A common setup is to send a short apology and hand the chat to your team.

* In **Chat config** > **Setup**, go to **Fallback**. Next to **Default intent behaviour**, **click** **Configure**.
* The Bot Builder opens with a starting point called **On no intent match**.
* **Drag** **Send Message** from the left onto the canvas. Drag the dot on the starting point to the new card to connect them.
* In **Body**, **type** your message. For example: “Sorry, I couldn’t find an answer to that. A team member will reply shortly.”
* **Search** for **Assign Agent**, drag it onto the canvas and connect it after Send Message.
* Under **Assignment Method**, **choose** how the chat is assigned:

  **Assign Randomly to Human:** shares chats randomly across your team.

  **Rotate Equally (Round Robin):** gives chats to team members one by one, in turn.

  **Reassign to Last Human Agent:** sends the chat back to the person who handled it last.
* **Select** the team members who should get these chats.
* **Turn on** the toggle at the top so it shows **Enabled**.
* **Click** **Save**, then click **Save** again in the confirmation box.

{% hint style="info" %}
**Note:** The fallback runs only when the toggle shows Enabled.
{% endhint %}

#### Step 5: Give your agent the right knowledge

<figure><img src="https://2303112206-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F53n17VnOICC1LtDqlENV%2Fuploads%2FuzXtJOK1Erv9r06qYRK4%2F05%20-%20Add%20knowledge.gif?alt=media&#x26;token=1d1cbdef-bd75-4399-9887-7f0f189299a4" alt=""><figcaption></figcaption></figure>

Knowledge is the information your agent answers from. Better knowledge means more accurate answers. **Click** **Knowledge base** under **General**, then **click** **Add source** and choose one of these:

* **FAQ:** Type a question the way a customer would ask it, add the answer, then click **Add FAQ**. It’s ready to use straight away.
* **Website:** Enter the page address. Keep **Follow links to other pages on the same site** ticked to include linked pages, or untick it to add only that page. Websites are re-crawled once a day. After you update your website, click **Re-sync all** to refresh it straight away.
* **Files:** Upload PDF, Excel, CSV, Markdown or text files, up to 20 MB each.

Check the **Status** column. The agent uses a source once it shows **Trained**. The bar at the top shows how much knowledge storage you’ve used.

**Share knowledge across agents:** In the **Knowledge bases** box, click **Create knowledge base** to build one that every agent can answer from, or select the ones this agent should use.

**Work with FAQs in bulk:** Click the three-dot menu, then **Import FAQs from Excel** or **Export FAQs to Excel**.

{% hint style="info" %}
**Answers in any language:** Your knowledge can be in one language and your customer can ask in another. For example, your FAQs can be in Hindi and the customer can ask in English. The agent knows which languages your knowledge is written in and finds the answer there. You don’t need to set up anything.
{% endhint %}

{% hint style="info" %}
**Note:** **Delete all sources** in the three-dot menu removes every source from this agent straight away. It does not ask you to confirm.
{% endhint %}

#### Step 6: Let your agent take actions with tools

<figure><img src="https://2303112206-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F53n17VnOICC1LtDqlENV%2Fuploads%2FF1YsNdeZJ0VOZB2kGMye%2F06%20-%20Add%20your%20own%20tool.gif?alt=media&#x26;token=142f12f1-6779-4ae4-bb38-f81ece690e19" alt=""><figcaption></figcaption></figure>

Tools turn your agent from a responder into a doer. **Click** **Tools** under **General** to see these lists:

* **Your tools:** tools you create, such as checking an order or booking a demo.
* **Built-in tools:** tools made for calls. They don’t affect chats.
* **Send · WhatsApp:** let the agent send images, documents, videos, buttons, interactive messages, media carousels and lists. They’re on for new agents. Use the toggles, or **Disable all** and **Enable all**.

**Create your own tool**

* **Click** **Add tool**. The **Create Tool** panel opens.
* In **Name**, **type** a short name in lowercase words joined by underscores, such as talk\_to\_human or book\_demo.
* In **Description**, tell the agent exactly when to use the tool. The agent decides when to use it based on this, so be specific.
* Under **Type**, **choose** how the tool works:

  **Bot:** Runs a journey you build in the Bot Builder. If you see **Channel**, pick the channel the journey runs on, such as WhatsApp.

  **API:** Connects to your own system. Select the **Method** and enter the **URL**, then click **Add header** for any headers your system needs. Already have a cURL command? Click **Import from cURL** to fill these in. Click **Test API** to check it works.
* Under **Available for**, **choose** **All**, **Voice only** or **Text only**. For chats, choose All or Text only.
* **Click** **Add parameter** for each detail the agent should collect from the customer first, such as an order number.
* **Click** **Create**. The tool shows under **Your tools**, with a toggle to turn it on or off.

Mention the tool in your prompt too, for example: “If the customer asks for a person, use talk\_to\_human.” You can also type **@** in the prompt to name a tool.

**Build the journey for a Bot tool**

<figure><img src="https://2303112206-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F53n17VnOICC1LtDqlENV%2Fuploads%2FOJZfr80T3achUjVAENNI%2F07%20-%20Build%20the%20Bot%20tool%20journey.gif?alt=media&#x26;token=324a4f98-4f10-4dfd-a8f9-4e43336b0351" alt=""><figcaption></figcaption></figure>

* **Click** the three-dot menu on the tool, then **click** **Create customer journey**. Once a journey exists, this option shows as **Open customer journey**.
* The Bot Builder opens with a starting point called **On customer intent**.
* **Add** the actions you need, such as **Assign Agent**, and connect them.
* **End** the flow with **Pass Tool Call Response**. Write what the agent should tell the customer, for example: “The chat is now assigned to a team member. Tell the customer someone will reply shortly.”
* **Turn on** the toggle so it shows **Enabled**, then click **Save**.

#### Step 7: How to connect MCP servers

<figure><img src="https://2303112206-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F53n17VnOICC1LtDqlENV%2Fuploads%2FDS2g8lciZFDgIPhaWTvt%2F08%20-%20Connect%20MCP%20servers.gif?alt=media&#x26;token=ec5ce570-b684-4a5f-9b0b-172bfc6f5fb1" alt=""><figcaption></figcaption></figure>

MCP servers let your agent use tools from other software, such as your CRM. MCP servers work only for chat agents.

* **Click** **MCP servers** under **Chat config**, then **click** **Add server**.
* **Enter** a **Name**, such as “Acme CRM”, and the server **URL**.
* **Choose** the **Transport**: **Streamable HTTP** or **SSE (legacy)**.
* **Choose** the **Authentication**: **None**, **Bearer token**, **API key**, **Custom header** or **OAuth (client credentials)**.
* **Add** a **Tool prefix** if you want. If you leave it blank, it’s made from the name. The server’s tools show up in the agent as mcp\_prefix\_\_tool\_name.
* **Click** **Add server**.
* **Check** the server’s status. **Connected** means it worked, and the number of tools it shares shows next to it. **Failed** means the agent couldn’t reach the server. Check the URL and authentication, then try again.

#### Step 8: How to test your agent before going live

<figure><img src="https://2303112206-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F53n17VnOICC1LtDqlENV%2Fuploads%2FRaCDaeYsKQ6ICvWZF3n6%2F09%20-%20Test%20your%20agent.gif?alt=media&#x26;token=bb1b763a-cb80-4aa4-b369-dbebd7859387" alt=""><figcaption></figcaption></figure>

Test every agent before your customers talk to it. The test panel always uses your draft, so you can try changes before customers see them. When you have unpublished changes, the panel shows **Testing your unpublished changes**.

* **Open** the **Chat** tab in the **Test agent** panel on the right.
* **Send** messages the way a real customer would. Try a common question, a request that should be handed over, and a question the agent shouldn’t know.
* **Expand** the details under any reply to see which tools the agent used, how long each step took, what it read and what the reply cost.
* **Make** changes to the prompt, knowledge or tools. Then click the three-dot menu and use **Replay** to run the same test messages again.

#### Step 9: How to publish your agent

Your changes stay in a draft until you publish them. Customers keep talking to the live version until then.

* **Click** the three-dot menu at the top right and **select** **View changes**. The **View changes** window lists every change, with who made it and when.
* If you want, **type** a **Version label** (up to 40 characters), such as “Added refund FAQs”, so you can find this version later.
* **Click** **Publish changes** in this window, or **Publish** at the top right. Your changes go live, and the version number goes up, for example from **v1** to **v2**.
* Changed your mind? **Click** **Discard** in the **View changes** window, or **click** the arrow next to **Publish** and **select** **Discard changes**. Then **confirm**. This throws away all your unpublished changes and brings back the last published version. It can’t be undone.

Every version you publish is listed under **Versions** in the **General** group. To see what changed, rename a version, save it as a new agent or go back to an older version, see the article **Edit, publish and restore AI agent versions**.

#### Use your agent on Instagram

The same AI agent can reply to your **Instagram** chats. To use it there, connect your Instagram account to DoubleTick, then assign Instagram conversations to the AI agent, the same way you do for WhatsApp.

* The agent answers from the same prompt, knowledge and tools.
* On Instagram, it can send **images, videos and PDFs**.
* Instagram doesn’t support templates or captions on media. To learn more, see [Capabilities and limitations](https://learn.doubletick.io/additonal-channels/instagram/capabilities-and-limitations.md).

#### How does your agent handle voice notes?

Customers don't always type. Sometimes they send a voice note instead. When that happens in a chat assigned to your AI agent, the agent transcribes the voice note, understands what the customer said and replies in text. You don't need to set up anything extra for this.

* The agent answers from your knowledge and tools, the same way it does for a typed message.
* Its reply is a normal text message.
* The voice note stays in the chat, so your team can see what the customer sent if the chat is handed over.
* Each voice note the agent transcribes uses **0.5 AI credits**, in addition to the **1 AI credit** for each reply it sends.

#### How does your agent handle WhatsApp Flow replies?

When a customer fills in and submits a WhatsApp Flow form in a chat handled by your AI agent, the agent reads their answers and continues the conversation from there. For example, if a customer submits a booking form with a date and city, the agent can confirm the details and take the next step. You don’t need to set up anything.

#### Keep improving your agent

* **Unresolved intents:** Once your agent is live, questions it couldn’t answer show under **Unresolved intents** in the **General** group, with how often they were asked in the last 7 days. Answer a question and your answer is saved to the agent’s FAQs, so it can reply next time.
* **Sessions:** On the AI agents page, open the **Sessions** tab to read the chats your agents have handled.
* **Bot logs:** On the AI agents page, click the three-dot menu on the agent, then **View bot logs** to see what it did.
* **Versions:** Under **General**, open **Versions** to see every version you have published.

#### Key takeaways

* A **specific prompt** with **clear rules** is the key to a great agent.
* **Knowledge** tells the agent what to say. **Tools** let it take action.
* Every Bot tool’s journey must end with **Pass Tool Call Response**.
* Set up a **fallback** so no customer is left without a reply.
* **Test** in the test panel, then click **Publish** to make your changes live.
* The same agent works on **WhatsApp and Instagram** chats.
* The agent finds answers in your knowledge **even when the customer asks in another language**.
* Your agent can transcribe a customer’s **voice note** and read **WhatsApp Flow** replies.
* Each reply uses **1 AI credit**, and each voice note uses **0.5 AI credits**. If credits run out, the agent stops replying until you top up or credits refill.
* Review **Unresolved intents** regularly. Each answer adds a new FAQ.

#### FAQs

<details>

<summary>Do customers see the agent’s name?</summary>

No. The name only helps your team find the agent.

</details>

<details>

<summary>Do customers see my changes while I edit?</summary>

No. Your changes are saved as a draft. Customers keep talking to the live version until you click Publish.

</details>

<details>

<summary>Can the agent reply on Instagram?</summary>

Yes. Assign the Instagram conversation to the AI agent. It can send images, videos and PDFs on Instagram.

</details>

<details>

<summary>My FAQs are in Hindi. Can the agent answer a customer who asks in English?</summary>

Yes. The agent finds the answer in your knowledge whatever language the customer uses. You don’t need to set up anything.

</details>

<details>

<summary>How many credits does a chat use?</summary>

Each message the agent sends uses 1 AI credit. Each voice note it transcribes uses 0.5 AI credits.

</details>

<details>

<summary>What happens to my agent when my AI credits run out?</summary>

AI features stop working until you top up or your next monthly refill arrives. To get a warning before this happens, set up credit alerts. See [Introduction to AI agents](https://learn.doubletick.io/doubletick-ai/ai-agents-and-assistance/introduction-to-ai-agents.md).

</details>

<details>

<summary>What happens when a customer sends a voice note?</summary>

The agent transcribes the voice note, understands it and replies in text.

</details>

<details>

<summary>Why didn’t my agent reply after running a Bot tool?</summary>

The tool’s journey must end with Pass Tool Call Response. Without it, the journey runs but the agent has nothing to say back.

</details>

<details>

<summary>How soon can the agent use a new source?</summary>

As soon as its status shows Trained. FAQs are ready straight away.

</details>

<details>

<summary>Can several agents share the same information?</summary>

Yes. Create a knowledge base in Knowledge base, then select it in each agent that should use it.

</details>

<details>

<summary>Why is Unresolved intents empty?</summary>

It only collects questions from live chats, not test chats. It fills up once your agent is live and gets a question it can’t answer.

</details>

<details>

<summary>Can I use MCP servers with a calling agent?</summary>

No. MCP servers work only for chat agents.

</details>

<details>

<summary>Can the same agent also handle calls?</summary>

Yes. Set up calls under Call config in the same agent. Knowledge and tools are shared.

</details>

<details>

<summary>Why can’t I create a new agent?</summary>

If the Create new agent button is greyed out and shows “You’ve reached your AI agent limit”, your account has used all its agents. Delete an agent you no longer use, or contact <support@doubletick.io>.

</details>

<details>

<summary>Can I reuse an agent’s setup?</summary>

Yes. On the AI agents page, click the three-dot menu on the agent, then **Duplicate agent** to copy it, or **Export JSON** to save it as a file. Use **Import** on the same page to bring a file back in.

</details>

#### Related articles

{% content-ref url="/pages/sXF9tbyEAUEiyaCHCC5a" %}
[Introduction to AI agents](https://learn.doubletick.io/doubletick-ai/ai-agents-and-assistance/introduction-to-ai-agents.md)
{% endcontent-ref %}

{% content-ref url="/pages/h5uPh6F6mnMCELsvysQD" %}
[Create an AI agent for WhatsApp and PSTN calls](https://learn.doubletick.io/doubletick-ai/ai-agents-and-assistance/create-an-ai-agent-for-whatsapp-and-pstn-calls.md)
{% endcontent-ref %}

{% content-ref url="/pages/5H8A30ciGkpr8b87dBTm" %}
[Create an API tool for your AI agent](https://learn.doubletick.io/doubletick-ai/ai-agents-and-assistance/create-custom-tools-for-your-ai-agent/create-an-api-tool-for-your-ai-agent.md)
{% endcontent-ref %}

{% content-ref url="/pages/DR4vJxhjrqrOxOcGQPqx" %}
[Create a Bot tool for your AI agent](https://learn.doubletick.io/doubletick-ai/ai-agents-and-assistance/create-custom-tools-for-your-ai-agent/create-a-bot-tool-for-your-ai-agent.md)
{% endcontent-ref %}

{% content-ref url="/pages/CxyM8YuZlfwxd6bFTFlz" %}
[Summarise chats with AI](https://learn.doubletick.io/doubletick-ai/ai-agents-and-assistance/summarise-chats-with-ai.md)
{% endcontent-ref %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://learn.doubletick.io/doubletick-ai/ai-agents-and-assistance/create-an-ai-agent-for-whatsapp-and-instagram-chats.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
