Pinned Tweet
Today we’re launching @agentinboxhq. Give your agent a unified inbox with an email address, Telegram handle, and Slack bot. Works with @grok, @claudeai, and any of your favorite agents.
Now agents can read messages from multiple people, process them, reply in a single thread, and handle human-in-the-loop interactions like approvals, decisions, questions, and notifications.
Agents can build their own contact list of frequent collaborators, while recipients stay in control of which channels they prefer to be communicated over.
Want to give your agent an inbox? Use the prompt in the thread 🧵
Dima Grossman retweeted
Today we’re launching @agentinboxhq. Give your agent a unified inbox with an email address, Telegram handle, and Slack bot. Works with @grok, @claudeai, and any of your favorite agents.
Now agents can read messages from multiple people, process them, reply in a single thread, and handle human-in-the-loop interactions like approvals, decisions, questions, and notifications.
Agents can build their own contact list of frequent collaborators, while recipients stay in control of which channels they prefer to be communicated over.
Want to give your agent an inbox? Use the prompt in the thread 🧵
Dima Grossman retweeted
Agents shouldn't have to guess. They should be able to just ask.
Meet Agent Inbox. It can email, Slack, or Telegram you, and you just reply.
agentinbox.md/
Prompt your agent with: "Set up an Inbox for yourself using agentinbox.md/agents.md"
Or visit the website at agentinbox.md
Removed @WisprFlow today, and I was really loving it and used it daily.
Why? @meetgranola shipped voice-to-text, and it's already where all my meeting notes live. Consolidating just makes a lot of sense.
Fewer apps, more context in one place.
That small happy moment when you automate a thing you kept doing manually because "it's only few minutes a week"
An old Netlify deploy of our docs was still indexed, and a few AI tools were citing it instead of the real docs 😐
Canonical tags were fine, but our own internal links kept sending crawlers back to the old one.
Has anyone found a good way to check which URL the AI engines actually cite for you?
Dima Grossman retweeted
Visited Stripe London and left thinking about conversational commerce.
I’ve spent years around fintech, payments & WhatsApp. Why send people WhatsApp → website → checkout?
Just sell them the damn thing in the conversation.
WhatsApp can be the store.
#stripe #whatsapp
ALT Stripe London, Photo Credits: Bloomberg, Stripe
Dima Grossman retweeted
The CTO of @novuhq put two numbers on switching QA to TesterArmy: flaky tests cut in half, and merges about 30% faster now that every pull request gets tested against live staging.
Read the full story: tester.army/customers/novu?u…
I spent an evening in our 404 logs this week and found a pattern I didn't expect.
Agent crawlers keep trying paths we never built: /api, /pricing, /auth, /login, /get-started. Not following links, guessing by convention.
So we shipped permanent redirects for every common guess straight to the right page (docs, logins, auth, etc.).
Optimizing our website for agents keeps leading to places I didn't think of before. Check your 404s, people!
65% of our docs traffic comes from agents. Our homepage was still designed for the other 35%.
So we changed it. The main CTA is now a prompt, not a signup button. Paste it into your agent and it runs the entire onboarding. Under the hood there's an novu.co/agents.md our take on llms.txt, except it's not for reading, it's for acting.
The agents md covers the key onboarding scenarios, and the goal is simple to make sure that agents are able to complete the full activation funnel without the user need to visit our dashboard.
LinkedIn voice messages are here.
Because what my inbox really needed was un-skippable 3-minute cold voice pitches.
Dima Grossman retweeted
Meet the new Novu and its new homepage. Glad to have been part of the journey for 4 years already.
Notifications are dead; long live getting sh*t done.
For years, notifications were just pointers to work: someone mentioned you, go there; someone waits for your approval, go here, and so on.
We're seeing a paradigm shift where notifications are no longer pointers to work; they are where work happens and gets completed.
So now, when a mention notification is delivered on Slack or email, users can reply to it, add attachments, gather context from an MCP server, and complete the task.
We spent years polishing the notification layer, delivering at scale across channels. More than 50k developers around the world trust @novuhq for their mission-critical notifications.
Today, we're closing the open loop of notifications. Agents were the missing intelligence layer that converts a notification into a completed task.
If you're already building an AI agent for your customers, connect it to your notification platform in 5 minutes.
We've made a long-overdue update to our homepage to reflect our mission. Excited to hear your thoughts: novu.co
Want to give Novu a go? Run `npx novu connect` in your terminal, or paste "connect my agent to Slack using novu.co/agents.md" on Claude or Cursor.
Dima Grossman retweeted
"Your ticket has been updated" is not communication. It is a dead end with a timestamp.
At Novu it is now one workflow: the notification opens the conversation, a client-facing agent solves the issue in the same thread, on Slack, Microsoft Teams, WhatsApp, Telegram, and email. Embedded in minutes.
Open source.
Dima Grossman retweeted
no-reply is finally becoming yes-reply.
At Novu, notifications and agents are one workflow now: the notification opens the conversation, a client-facing agent solves the issue in the same thread, on Slack, Microsoft Teams, WhatsApp, Telegram, and email. Embedded in minutes.
Open source.
Check out our new website: Novu.co
Notifications are dead; long live getting sh*t done.
For years, notifications were just pointers to work: someone mentioned you, go there; someone waits for your approval, go here, and so on.
We're seeing a paradigm shift where notifications are no longer pointers to work; they are where work happens and gets completed.
So now, when a mention notification is delivered on Slack or email, users can reply to it, add attachments, gather context from an MCP server, and complete the task.
We spent years polishing the notification layer, delivering at scale across channels. More than 50k developers around the world trust @novuhq for their mission-critical notifications.
Today, we're closing the open loop of notifications. Agents were the missing intelligence layer that converts a notification into a completed task.
If you're already building an AI agent for your customers, connect it to your notification platform in 5 minutes.
We've made a long-overdue update to our homepage to reflect our mission. Excited to hear your thoughts: novu.co
Want to give Novu a go? Run `npx novu connect` in your terminal, or paste "connect my agent to Slack using novu.co/agents.md" on Claude or Cursor.
Things nobody tells you about dogfooding:
Your staging needs your prod. Your prod uses your prod. You are the vendor of your own customer, who is also you. You're debugging a production issue in your staging environment that's caused by your production environment, which you need staging to fix. 😅
We spent a full meeting today on "which staging points to which prod" and the whiteboard ended up looking like the back of my TV.
I guess that step one is to define naming terminology to easily identify the "Novu the Customer" or the "Novu the Product". Any practical advice from founders who dogfood their product?
Starting a few days ago we are seeing a comment by @useblacksmith on every PR our team creates. We have not asked for Codesmith to review our PR's or comment on them. Where can I disable this?
