mailer-gpt

MailerGPT

Feature: mail

MailerGPT integrates with the Nylas email API to read, parse, and send emails autonomously. It extracts relevant information from incoming messages and generates personalized outgoing email content based on user-defined goals.

What MailerGPT Solves

Automating email workflows typically requires complex IMAP/SMTP integrations with fragile parsing logic. MailerGPT uses the Nylas unified API to abstract email providers and the LLM to handle content understanding and generation, turning natural language goals into fully composed emails.

Enabling MailerGPT

MailerGPT requires the mail feature flag and three Nylas environment variables:

cargo install autogpt --features mail,gem

export NYLAS_SYSTEM_TOKEN=<your_nylas_system_token>
export NYLAS_CLIENT_ID=<your_nylas_client_id>
export NYLAS_CLIENT_SECRET=<your_nylas_client_secret>

Follow the Nylas setup guide to obtain these credentials.

How It Works

1. MailerGPT connects to the Nylas API using the configured credentials
2. It reads emails from the configured inbox and parses their content
3. Based on the agent's behavior goal, it composes reply or new outbound email content using the LLM
4. It sends the generated email via the Nylas API

SDK Usage

use autogpt::prelude::*;

#[tokio::main]
async fn main() {
    let persona = "Customer Support Agent";
    let behavior = "Read incoming support emails and draft polite, helpful replies.";

    let agent = MailerGPT::new(persona, behavior).await;

    AutoGPT::default()
        .with(agents![agent])
        .build()
        .expect("Failed to build AutoGPT")
        .run()
        .await
        .unwrap();
}