←
Nylas7,479 posts
Nylas avatar

Nylas

@nylas

Connect to any inbox or calendar. The API that lets you integrate with Gmail, Microsoft, IMAP, Zoom, and 250+ mail, calendar and meeting providers in 5 minutes.

◷ Joined April 2013▧ 1,094 Photos and videos
Nylas avatar
Nylas@nylas·Jun 17

Agent Accounts is generally available today. 🚀 Your AI agent gets its own hosted email and calendar identity, provisioned with a single API call on your domain. Send, receive, and schedule, with policy enforcement and rate limits built in. nylas.com/products/agent-acc…

Post attachment
Nylas avatar
Nylas@nylas·Nov 17

A calendar invite from a stranger can carry instructions your AI agent reads the next time it checks your schedule. Clean the content, label it as untrusted, and limit what the agent can send. Read how to defend an agent against prompt injection: nylas.com/blog/email-prompt-…

Post attachment
Nylas avatar
Nylas@nylas·Nov 13

Pour one out for everyone who spent yesterday explaining it wasn’t their code.

Post attachment
Nylas avatar
Nylas@nylas·Nov 12

That’s a wrap on Web Summit. Thank you to everyone who stopped by to show us what you’re building with email, calendar, and AI agents. If we missed you in Lisbon, the product flyers are still at nylas.com/websummit/, and we’d like to hear what you’re working on.

Post attachment
Nylas avatar
Nylas@nylas·Nov 12

The Nylas Shared Google App is already validated at AL2, the highest CASA level, so you skip Google's verification queue and the annual security audit. Read about how to get Google OAuth approved for your org: nylas.com/blog/google-oauth-…

Post attachment
Nylas avatar
Nylas@nylas·Nov 9

You can’t predict what an agent will decide. You can set what it’s able to reach. With Nylas IAM, a read-only key for one mailbox means a send attempt returns 403 and lands in Access Activity. nylas.com/blog/ai-agent-perm…

Post attachment
Nylas avatar
Nylas@nylas·Nov 5

Nylas is heading to @WebSummit in Lisbon, Nov 9–12 🇵🇹 Find us at Booth E493, Pavilion 4. Tell us what you’re building, whether that’s email and calendar in your product or an inbox for your AI agents. Stop by to enter our Bose QuietComfort headphones giveaway 🎧

Post attachment
Nylas avatar
Nylas@nylas·Nov 4

58% of teams with customer-facing AI agents let them email customers as an employee at least some of the time. When the agent gets it wrong, the customer blames that employee and the audit log shows their name. New data from our H2 2026 survey: nylas.com/blog/customer-faci…

Post attachment
Nylas avatar
Nylas@nylas·Oct 28

Overjoy gives 1,000+ CPG brands an AI sales team, working from each founder’s inbox. Four engineers keep it running. The Nylas Email API and Hosted Authentication handle the inbox connections. How they built it: nylas.com/case-study/overjoy…

Post attachment
Nylas avatar
Nylas@nylas·Oct 27

Companies trust their AI agents less than they did in December. They’re giving them more freedom anyway. Full trust fell from 37% to 32%; no-approval autonomy nearly doubled. We compared both waves of our survey question by question: nylas.com/blog/more-rope-few…

Post attachment
Nylas avatar
Nylas@nylas·Oct 23

Can your team produce a full audit log of what your AI agents did last week, and say which agent did what? 53% of 1,030 respondents can. 37% partially. 10% can't or don't audit at all. nylas.com/agentic-ai-report-…

Post attachment
Nylas avatar
Nylas@nylas·Oct 20

When something does go wrong, nearly half of teams cannot say which agent did it. Only 53% can produce a full audit log of what their agents did last week.

Nylas avatar
Nylas@nylas·Oct 20

The incident rate rises with scale. 45% of orgs running 1 to 5 agents reported an incident. 70% of those running more than 20.

Nylas avatar
Nylas@nylas·Oct 20

The most common failure: an agent triggered a workflow the team had to reverse or correct (31%). A quarter had an agent update the wrong record. 18% had one send an incorrect or inappropriate email.

Nylas avatar
Nylas@nylas·Oct 20

54% of companies had an AI agent make a mistake in the past 90 days. We asked 1,030 people what their agents did wrong last quarter. The State of Agentic AI: H2 2026 is out today, ungated. nylas.com/agentic-ai-report-…

Post attachment