cool product
but i would like to add some point.
so sharing only agent.md is not sufficient i feel try if you can copy a session of claude and import it into opencode/codex.
tool that track session, i dont mean every message i mean useful context.
behind the scene agent is using sqlite or some other their format is it not that much hard to reverse engineer.
Very cool. I just posted something similar about sharing markdown and it's interesting to see how you've taken it further.
I wonder how much the providers will build this in as a core feature - I'm sure gemini will have shared artifact workspaces at some point.
Have you thought about diffs and versioned documents? I'm curious waht's out there adn couldn't find a good solution but I think collaborative ai-enabled editing is in its infancy.
Versioned docs are already supported - each publish saves a new version with the previous ones remaining accessible. Diffs are coming too.
I think the co-editing/collaboration is very interesting. We currently support comments by anyone in your org with agents editing and resolving based on them (importantly, keeping the identity on whose behalf they’re acting intact).
I think all major platforms will add some sort of artifact sharing (case in point Anthropic today).
Where we differ is in thinking that agent knowledge work needs to be on a neutral surface, never vendor locked, and accessible to everyone in your org, regardless of whether they hold a certain vendor’s license. Hence our take on flat org-based pricing (no per seat) in the paid tiers.
As mentioned in another comment, I believe that an agent-neutral surface needs to exist for agent produced knowledge work, and even more importantly so if everyone in your org needs to have access.
Knowledge work shouldn’t have vendor lock-in, since the best model tends to change every two months, people use multiple agents, and not everyone in your org has a license.
good idea - i built something similar for screen-recordings. I think tools that act as primitives within the harness are very early but potentially large tailwinds. BOL
It’s certainly a convenient default for heavy Claude users.
For multi-agent users, orgs that don’t have licenses for everyone in their org, orgs that don’t want to stay vendor locked, a neutral surface is still required.
That’s something I certainly still believe - agent agnostic surfaces are necessary for knowledge work.
cool product but i would like to add some point. so sharing only agent.md is not sufficient i feel try if you can copy a session of claude and import it into opencode/codex. tool that track session, i dont mean every message i mean useful context. behind the scene agent is using sqlite or some other their format is it not that much hard to reverse engineer.
Very cool. I just posted something similar about sharing markdown and it's interesting to see how you've taken it further.
I wonder how much the providers will build this in as a core feature - I'm sure gemini will have shared artifact workspaces at some point.
Have you thought about diffs and versioned documents? I'm curious waht's out there adn couldn't find a good solution but I think collaborative ai-enabled editing is in its infancy.
Thanks!
Versioned docs are already supported - each publish saves a new version with the previous ones remaining accessible. Diffs are coming too.
I think the co-editing/collaboration is very interesting. We currently support comments by anyone in your org with agents editing and resolving based on them (importantly, keeping the identity on whose behalf they’re acting intact).
I think all major platforms will add some sort of artifact sharing (case in point Anthropic today).
Where we differ is in thinking that agent knowledge work needs to be on a neutral surface, never vendor locked, and accessible to everyone in your org, regardless of whether they hold a certain vendor’s license. Hence our take on flat org-based pricing (no per seat) in the paid tiers.
I like the idea .. and believe there’s need for it beyond what Claude or ChatGPT can implicitly offer. Good luck!
Thank you!
As mentioned in another comment, I believe that an agent-neutral surface needs to exist for agent produced knowledge work, and even more importantly so if everyone in your org needs to have access.
Knowledge work shouldn’t have vendor lock-in, since the best model tends to change every two months, people use multiple agents, and not everyone in your org has a license.
good idea - i built something similar for screen-recordings. I think tools that act as primitives within the harness are very early but potentially large tailwinds. BOL
Did you see the Claude Artifacts that was just released?
Yep, saw it.
It’s certainly a convenient default for heavy Claude users.
For multi-agent users, orgs that don’t have licenses for everyone in their org, orgs that don’t want to stay vendor locked, a neutral surface is still required.
That’s something I certainly still believe - agent agnostic surfaces are necessary for knowledge work.