Hey everyone, another question from the Velozity team as we get closer to launch.
Last time we asked how AI work moves between people on a team. This time we want to dig into trust.
Right now most people keep their AI on a short leash: it drafts, it suggests, but a person clicks send. That's safe, but it also means every handoff between a person and their AI, or between two people's AIs, still needs a human in the loop to move things along.
We're building Velozity so agents can work more directly inside a team's shared space, chat, tasks, files, calendar, with a person approving anything that leaves the workspace rather than approving every single step along the way.
The part I like is that agents sit alongside my teammates, so I don't have to move work between tools by hand.
@shreya_verma12 Thank you, Shreya! That's the part we care about most. Moving work between tools by hand is a hidden tax on every team, and it disappears once agents sit in the same place as your teammates. Which handoff has saved you the most time so far?
@shreya_verma12 Adding to Ishu's note: the reason it works is that the agent already has the context your teammates have, so there is nothing to copy over or explain again. A decision made in a call is already where the agent picks up the task. Curious which hand-off you would like us to make smoother next.
Support from a fan of the build. I watched the team work on this for a long time, and it shows in how polished it feels. Well done.
@shivansh_singh26 Thank you so much, Shivansh, that means a lot. It took a lot of patience from the whole team, and comments like yours make it worth it. If you get a chance to try it today, I'd love to hear what you think works and what doesn't.
@shivansh_singh26
Shivansh: having someone who watched the build say it feels polished means a lot to all of us. Most of that polish came from using Velozity ourselves every day and fixing whatever slowed us down. If anything feels rough when you try it, tell us plainly, that feedback is what keeps it improving.
Sounds awesome! Also, amazing editing of the video! Congrats!
@lakshya_singh Thank you, Lakshya! 🙌 Not everyone will see the video the same way, and that's fine, but I'm really glad it clicked for you. We wanted it to show real teammates doing real work, and your note tells me it landed.
@lakshya_singh Lakshya: thank you, the video took a lot of care and it's great to hear the editing stood out. If you try Velozity after watching it, tell us whether the product feels the way the video promised. That's the comparison we care about most.
This looks cool! Congrats
@louislecat Thank you so much, Louis! Please give it a spin. I'd be happy to give you three months of Pro on us. Just sign up and drop me a note here, and I'll switch it on. Would love to hear what you think once you've tried it with your team.
@louislecat
Thanks Louis, means a lot coming from a fellow maker. Ishu's offer stands, and if you do try it with the Upstream team, the quickest way to see the point is to run one real meeting through it and watch the follow-ups turn into tasks. Honest feedback welcome, including what did not land.
@bhavyapatel Thanks so much, Bhavya! Really glad you were with us from the early days. Your support means a lot to the whole team.
Congrats on the launch, I really like the User interface, One question about "use your existing AI subscription".
Anthropic's rules for Claude subscriptions are strict, so I want to understand how this works before I connect mine:
Do you run the official, unmodified Claude Code, or your own client?
Does the login happen on Anthropic's own page, and do you ever store my Claude token on your servers?
On the cloud computer, where is my Claude login saved?
Can teammates use an agent that runs on my Claude subscription?
Thanks, a clear answer here would help many users feel safe.
Thanks, @navid_shad. Fair questions.
Official Claude Code, unmodified. We don't ship our own Claude client. Velozity runs the Claude Code that Anthropic distributes. If you already have it installed, we use yours. If you don't, we fetch it with Anthropic's own installer from Anthropic's site, so updates stay on their channel. We start it in its standard non-interactive mode with its documented options.
You log in on Anthropic's page, and we never store your token. "Connect Claude" runs Claude Code's own sign-in, which opens Anthropic's login page in your browser. The token is saved by Claude Code itself, in the same place it uses when you sign in from a terminal. Velozity never reads, stores, or refreshes it, and it is never sent to our servers.
On a cloud computer, the login is saved only on that computer. It is the same official sign-in, run on your private cloud computer, and Claude Code saves the login in its own file there. It is not copied into our database. When you stop the computer, it saves its disk as a whole, so your sign-in is still there when it starts again.
No, teammates cannot use an agent that runs on your subscription. Only the person who owns an agent can run it. A teammate in the same Space can see its replies, but if they try to trigger it, the server refuses. Each person connects their own Claude account for their own agents.
@ishudarshan That's an interesting architecture that users bring their agents on behalf of their self and and each user just can trigger its own agent.
What about the runtime environment, where agents get run online or local? this part has been mention in the website but honestly still unclear for me:
1. Do I get a vm from Velozity?
2. Do Velozity runs the agents on my local machine?
@navid_shad Both, and you choose per agent.
Yes, you can get a cloud computer from Velozity. It's a full Linux virtual machine that belongs only to you, not a shared container. It runs on boat (formerly Box) by ASCII, whose machines and their saved disks are hosted in the EU. Your agent works there even when your laptop is closed; you can watch its screen live, and it stops itself when idle and picks up where it left off.
Yes, agents can also run on your own machine. In that case, Velozity starts the Claude Code or Codex already installed on your computer, using your login and your files. Nothing is moved to a server to run it, and it works while the Velozity app is open.
Same rule in both cases: only you can trigger your own agent.
@ishudarshan appreciate your answers
@navid_shad
Thanks Navid, fair questions. We run the official, unmodified Claude Code, and you sign in through Claude's own login on Anthropic's page. We never see or store your Claude token on our servers. On a cloud computer the login is saved on that computer itself, in the same place Claude Code normally keeps it. Teammates can see your agent's work in a shared Space, but only you can run it, so your subscription is never shared.
@pawani_krishna Thanks, Pawani! Yes, it works well for study groups. Your group's chats, notes, and AI help all live in one shared space instead of being scattered across apps. There's a free plan to get started, so you can try it with your group at no cost.
@pawani_krishna
A good way to start is one Space per subject or study group. Drop in your notes and lecture material, and everyone in the group, plus any AI agent you add, works from the same set. If you record a group call there, the discussion stays in the Space too, so nobody has to rewrite what was covered. Would love to hear how it goes for your group.