> For the complete documentation index, see [llms.txt](https://documentation.pushly.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://documentation.pushly.com/platform/ai-and-agents.md).

# AI & Agents

The Pushly MCP (Model Context Protocol) server is a secure, read-only connection that lets AI tools like Claude and ChatGPT access your Pushly data -- notifications, segments, feeds, journeys, subscriber acquisition, and performance insights -- to answer questions, surface trends, and analyze results in natural language.

### What is Model Context Protocol?

Model Context Protocol, or MCP, is an open standard that lets AI agents connect to and query data from external platforms. It has two parts:

* **MCP client:** The AI tool where you interact -- such as Claude or ChatGPT.
* **MCP server:** A service provided by another platform -- like Pushly -- that defines which data and tools the AI can access.

When you connect an MCP client to Pushly, the AI can answer questions about your data directly, without you having to export anything or manually interpret dashboards.

### What you can do with the Pushly MCP

The Pushly MCP is read-only. AI agents can read your data and surface insights, but they cannot create, edit, or delete anything in Pushly.

**Notifications** -- Ask which notifications performed best, compare click rates across sends, or pull uniques for a specific notification.

**Segments** -- Look up segment definitions, check audience sizes, or identify which segments are in use across your notifications.

**Feeds** -- Review feed configurations and preview content to understand what a feed is currently returning.

**Permission prompts** -- Check prompt configurations and pull subscriber acquisition data for a domain or date range.

**Journeys** -- Ask about enrollment counts, step-level performance, conversion rates, and decision branch breakdowns.

**Organization insights** -- Get notification performance trends and subscriber acquisition data across all domains in your organization in one query.

### Who it's for

The Pushly MCP is useful for anyone who wants to query their notification data through an AI tool rather than navigating dashboards. Common use cases include:

* Reviewing performance across a date range without building a report
* Comparing segments or notifications in natural language
* Pulling journey and acquisition trends for a weekly or monthly check-in
* Giving an AI assistant context about your Pushly setup so it can help draft strategies or identify patterns

### Connecting an AI agent

To connect an AI client, connect using the Pushly MCP server URL. The connection uses OAuth -- you sign in to your Pushly account, approve the permissions you want to grant, and the client is authorized.

Access is tied to your Pushly user account. If your account is No Access, you cannot connect an AI agent. Permission changes on your account apply immediately to any connected client.

**Supported clients include:** Claude, Claude Code, ChatGPT, Cursor, and any MCP-compatible client that supports OAuth remote server connections.

For step-by-step setup instructions, see:

* [Connecting Claude](https://documentation.pushly.com/platform/ai-and-agents/connecting-claude-to-pushly.md)
* [Connecting ChatGPT](https://documentation.pushly.com/platform/ai-and-agents/connecting-chatgpt-to-pushly.md)

### Permissions

When you connect a client, Pushly asks you to approve six permission categories. You can approve all six or only the ones you want the AI to access.

| Permission                 | What the AI can read                                  |
| -------------------------- | ----------------------------------------------------- |
| `domain:read`              | Domain configuration and settings                     |
| `domain_notification:read` | Notification content and performance                  |
| `domain_segment:read`      | Segment definitions and audience sizes                |
| `domain_prompt:read`       | Permission prompt configurations and acquisition data |
| `domain_feed:read`         | Feed configurations and content                       |
| `domain_campaign:read`     | Journey configurations and performance metrics        |

You can revoke a client's access by disconnecting it within your AI client.

### Frequently asked questions

**What Pushly data can an AI agent access?**

Only data covered by the permissions you approved during setup. The agent can read notifications, segments, feeds, permission prompts, journeys, and performance data -- but it cannot access subscriber-level personal data.

**Can an AI agent change my Pushly data?**

No. The Pushly MCP is read-only. No connected AI agent can create, edit, or delete anything in Pushly.

**What happens if my Pushly role changes?**

Permission changes apply on the next request. If your account is deactivated or changed to No Access, the connected client loses access immediately.

**Can I connect multiple AI clients at the same time?**

Yes. Each client goes through its own OAuth flow and maintains its own connection. You can connect Claude, ChatGPT, Cursor, and other clients simultaneously and manage each connection separately within your AI client.

**Does using the MCP count toward any API limits?**

MCP requests are separate from the Pushly REST API. They do not consume REST API quota.

**What if I only want the AI to access some of my data?**

Approve only the permission categories you need during the OAuth flow. You can update your permissions at any time by disconnecting and reconnecting.

**Is subscriber personal data ever returned?**

No. Tools that would return subscriber-level personal data are not available. The MCP returns aggregated metrics, configuration data, and content -- never subscriber identifiers, emails, or other personal information.

### Legal notice

The Pushly MCP server receives requests from your AI client exactly as that client's underlying model formulates them. When you type a question in natural language, the AI model interprets it and translates it into one or more tool calls to Pushly. Pushly receives and executes the tool call as sent.

When Pushly returns data, that response is sent back to your AI client, which then interprets, formats, and presents it to you. Pushly does not control how the AI model presents, summarizes, or describes returned information.

Always verify AI-generated summaries against your data in the Pushly dashboard before using them to make decisions.


---

# 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://documentation.pushly.com/platform/ai-and-agents.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.
