Praxos is a messaging platform with built-in company memory
Praxos offers people and AI agents a place to talk and work together. It is a messaging platform that remembers the context around a conversation: who said what, when they said it, what people were trying to do, and what happened afterward.
That context can then be used by the next person, AI agent, or even you, a week later. You can pick work back up without needing to get hold of another person to explain things again for you… or give you a refresher.
A surprising amount of work exists only because one person knows something another person needs, or because we need to coordinate activities with another person.
Things like:
“What did we tell Jane last week?”
"What was I doing on this Claude session yesterday?"
"John, can you please send me the proposal we sent ACME on Monday?"
"Is anyone else taking care of the client's request?"
"I'll send you a calendar invite to review next year's roadmap and finances after the call."
You're in good hands if you'd rather never deal with these again. We think AI can handle much of this for you.
Sometimes, AI should take care of these without you having to ask at all. When you do ask, you should get the answers you need or work moving without waiting on another person to see your message.
That's how we can give you back your time, and more room to do the things only you can do.
Why now?
Because AI is multiplying the number of conversations any one person has at a given time.
Think about it: for most people, the main entry point to using AI is messaging.
So, you now have to add the number of conversations you have with AI tools like ChatGPT and Claude to all the other conversations you were having with your coworkers. Each of these conversations holds valuable bits of information that the others don't. Someone has to move the context from one place to the other.
This is an old problem with a new dimension.
Before language models, you already had to summarize things for other people, forward messages, or track down the person who knew something you needed. Now, you might find yourself doing the same thing across different AI chat sessions.
AI created more conversations for us to keep track of, and we believe that AI can also help connect them.
Our theses
We believe that, though Praxos as a product will evolve over time, a few principles will remain constant. Rather than give you a feature list of what we want to build, we want to share the ideas that guide us.
1-What one person or agent learns should be useful to the next
It takes a lot of effort to share what we know.
There is often a great deal of context behind why a decision was made, or why something works the way it does.
Across many teams there is that one person that's a reference point to the rest. Wisdom, by way of the decisions, explanations, mistakes, and experiences they've lived remains effectively locked in their head until you ask. We want to make that knowledge useful to to you, whenever you need it.
Likewise, AI creates another version of the same problem. Model sessions are becoming new repositories of knowledge. You might spend one hour working on a problem within three different Claude sessions. What one session learns should become clay that other teammates, human and AI alike, can work with later.
Our business is making sure that knowledge remains useful after the conversation where it was created has ended.
2-Software should update itself as work happens
Until now, a lot of business software has existed because people had to translate what happened in the real world into records that someone, or some computer, could understand later.
Coordination has a cost. It also turns out that a surprising amount of software is just places to record it.
Think about all the small tasks "getting work done" includes when you're involving other people, or even your future self:
Updating a CRM every time you have a conversation.
Creating a task on some project management tool after you agree on doing something.
Sending a calendar invite when you agree to a meeting.
Holding weekly sync meetings to make sure everyone is on the same page after a week of work.
The list goes on.
We are at a point where technology can understand much of what is happening while people and agents talk and work.
A lot of this coordination should happen without you putting in the effort towards making it happen. Software should be able to notice when something still needs to happen.
If you promised to send a proposal tomorrow, the useful thing is not just to remember the promise. It is to help make sure the proposal gets sent.
We want to make that true.
3-Multiplayer AI involves using agents, but durable agents should remain rare
A lot of work is temporary. You should be able to say what needs to be done and let the system figure out what resources it needs.
Most agents we use for this work should be temporary. Why should we even know they exist? The valuable bits are in the work they do and the knowledge they leave behind.
A world where we treat every agent like a unique member of the team, with its own identity, knowledge, permissions, and abilities, is a nightmare. People management problems that plague organizations with 10,000 employees suddenly expand to every company in the world.
Consider a startup with 5 people, where each person brings 3 agents. You're now managing 20 entities across chats!
We don't want that. Do you want it?
Instead, we propose revisiting the basics of what an agent is: a model + context + tools.
Of these, only context needs to survive. But most context is useful only sporadically. Picking what matters and making sure it reaches the right person or model is tricky.
That is why we care so much about making sure that context, knowledge, are easy to share. Agents are just a tool.
Boundaries
Praxos sits near several familiar categories. To avoid risk of confusion, we lay ahead our product boundaries.
Not a wrapper around Slack, Whatsapp, Telegram or Discord. Messaging and calling are fundamental parts of Praxos.
We’ll maintain connections to Slack, Whatsapp, Telegram, SMS and Discord for as long as you have teammates who refuse to leave. But these are bridges. They are supposed to help you get value quickly and to transition without hiccups.
Praxos gets better for you if we can get you and your team over here.
Not a "company brain" or another place to manually maintain company knowledge. We believe that there is little value in organizing knowledge for its own sake. Archiving work is a chore. Let it fade into the background.
A cabinet of neat, organized files remains gathering dust until someone has a need to search for a file. We view company brains and most repositories of knowledge the same way. Value appears only when someone needs something inside it, but the repository itself is not valuable.
We believe that it's more convenient, and easier, to organize context in the background and make it available to you whenever you, or your agents, need it.
Not an ecosystem that's dependent on our agents. We enable communications and build the infrastructure that enables it. We don't want to build every agent your company uses.
You may use Claude today, Codex tomorrow, an internal agent next week, and something that does not exist yet a year from now. Praxos should work with any of them.
We are building Praxos to be extensible so that our agents, your agents, and agents built by other companies can all work with the same context.
Who is Praxos for?
Praxos is an ideal fit for teams who frustrated by existing messaging platforms and their developer tooling. It works best for early-stage companies ranging from a solo founder to teams of 25 people.
If you live between Slack, Claude, Codex, customer conversations, docs, and email… relaying what you learn from one conversation to another is starting to get overwhelming, then Praxos is for you.
Getting started
You have a few options.
If you're a builder looking to try it on their own using a CLI, paste sh -c "$(curl -fsSL "https://app.mypraxos.com/claude-code/install.sh?$(date +%s)")".
If you would like to talk to a person to get guided setup, book your call at https://cal.com/lucas-urbisaia-u4ip5c/30min.
And, if you’re starting a new company and just to pick a messaging platform, reach out! We’ll onboard you within the day. Ping Lucas at @lmurbisaia on X or text him at +1-929-271-7338.