Composer - Multiplayer markdown for you, your team, and your agents.
by•
Composer is a real-time, multiplayer markdown editor where people and agents can work side-by-side. Instantly share markdown generated by your agent with teammates, edit in real time, leave comments, suggestions, and share context. Your agents join as true collaborators, working directly alongside you and your team.
Replies
Composer
Hey Product Hunt 👋 I’m Jesse, co-founder of Composer.
Agentic workflows today are mostly focused on written documents. These same workflows also feel like they are mostly solitary. The hard part isn't writing a document, it's shaping it *together*. Our agents write everything, including the wrong things. Then, when you’re ready to share what you’ve done with others, you have to copy results over to a shared space, sending screenshots, committing to a repository, or chatting about what you did. There’s no straightforward way for people and agents to work together in real-time.
That’s why we created Composer.
How it works:
- Composer is all Markdown under the hood.
- Markdown docs can be shared directly from your coding agent session.
- Agents have full document access via MCP and can reply to comments and post suggestions.
- Supports real-time collaboration and access controls with both people and agents. Document versioning is coming soon!
Here’s how teams are already using Composer:
- Quickly shape a plan together before you start implementation.
- Use comments and suggestions to quickly iterate with your agent on document content.
- Easily share and collaborate in Markdown without losing shared context.
Check it out at: https://usecomposer.md
We can't wait to hear your feedback! We're working hard to iteratively polish the product and your perspectives will help us make Composer the best it can be.
Let us know what you think!
Jesse
@nottil The idea of agents becoming actual collaborators instead of just content generators is really interesting. How do you handle conflicts when multiple people and multiple agents are editing the same document at the same time?
Composer
@nottil @nicole_hynek Hey Nicole! Technical co-founder here. We use CRDT data structures behind the scenes to ensure the latest changes are seamlessly merged in real time. This applies to both human and agent collaborators via MCP.
@nottil @josh_philpott That's really interesting. Using CRDTs feels like the right foundation if you're serious about humans and agents collaborating in the same document rather than taking turns.
I think seamless conflict resolution becomes even more important once multiple agents are involved, so it's great to see you've thought about that at the infrastructure level from the start. Awesome work on this!
Composer
@josh_philpott @nicole_hynek Thanks Nicole, we couldn't agree more!!
Love the "Get started" flow right into the doc from the website.
Composer
@heyderekj Thank you! its been fun to work on little animations to create the feeling of continuity
the gap between 'ai generated this document' and 'now the team needs to edit it together' is where everything gets messy. copy paste into google docs, lose formatting, lose context. agents and humans editing in the same doc in real time makes sense
Composer
@tina_chhabra Yes! This is exactly what Jesse and I set out to solve when we started Composer. Generated markdown lives on machine and it's clumsy to work on with teammates. We wanted to make it absolutely seamless to share and collaborate on.
The "and your agents" part is what I want to understand better. Most multiplayer doc tools bolt on an AI layer that can read and maybe append, but real agent collaboration means handling conflicts when an agent edits something a human is mid-way through. How does Composer resolve that? And are agents operating on the same CRDT or OT layer as human cursors, or is agent writes handled separately and merged in?
Composer
@fberrez1 Hi Florent! Yes, agent collaborators operate on the same CRDT layer as humans.
Agent collaboration gets interesting when suggestions and comments carry the agent’s assumptions.
In a shared doc, I’d want each agent edit to answer: what source/context did you use, what did you infer, and what should a human verify? Real-time CRDT solves edit collision, but teams will also need confidence/provenance collision when humans and agents disagree.
Are you planning per-suggestion source trails or decision history as versioning lands?
Composer
@jim_jeffers Yes, totally agree. Provenance tracking, document versioning, and grounding are all currently top of mind as we move forward.
Freeplay
Very cool, I've been needing a tool like this recently! Too much text to write and neither markdown files in GitHub or Google Docs are the right way to do it.
Composer
@nicoto Thank you so much for the support, Nico! We couldn't agree more and are excited to hear how you like it and what you would improve about the experience to help you accomplish your goals!
Composer
@zaid_mallik1 Still a little early to notice trends here, but we're definitely interested in seeing how teams use Composer and if information makes sense to be structured differently or not in these different contexts. Curious to hear more if you try with more folks!
mailX by mailwarm
How do you handle the history and undo section when both humans and agents are editing at the same time?
Composer
@othman_katim This is very much on our radar!
Currently we are using CRDT undo stacks to handle this but we know that is not a good solution in the medium term. The next thing we're focusing on is a more robust versioning system.
Congrats on the launch! I think it's super interesting to collaborate with your agent within the doc itself. It's been surprisingly useful to just send a .md to composer and brainstorm against a specific paragraph. This seems to be a good way to collab on an org wide skill.
Composer
@maiprofile Thanks Mai!! Really appreciate the support. Oh we'd love to see and hear more about how Composer can help you collaborate on org-wide skills - that's a great usecase!