I still believe this space to be unclaimed, and I am yet to see who's going to crack the demand and needs of the niches out there.
I don't think we even have a great understanding of the market, the niches, and where it is going.
It's also evolving very aggressively constantly, sticky workflows get replaced dramatically whenever new UX or tooling brings a shift into how things can be done.
I'm very curious if anyone has good research or idea of this space mapped out somewhere, I am trying to build for it too but it's really foggy right now and the North Star moves often.
Love these threads for sharing experience. Using Orca has been a game changer (which I found and started using because I saw somebody using Conductor)
These sorts of management tools really speed up development by automating all the PR management, such as making new worktrees, etc. and then super speed up the multitasking between different parallel feature development.
Excited to see if this one works better for me, and to try all the new alternatives in these comments.
There's also sometimes options that try to have a tighter integration with a single harness, which honestly seems to be better cause ACP is fairly bare:
Personally, I found Paseo to be very pleasant, except sometimes I had sub-agents kinda hang.
Kepler had a really nice issue tracking integration (GitLab/GitHub Issues, Trello, Linear, Jira) but their UI felt like vibe coded slop, like the core experience doesn't feel very good, but they keep adding shit that nobody needs like a graph visualization of what the agents are doing, instead of letting me drag the tasks around the Kanban board myself or tell the agents how to do that and how to transition the issue statuses. Hopefully it's gonna get better within a few years, still early on in development. Still, the idea of an issue tracker integration is brilliant and all harnesses should consider supporting that, e.g. pick which tickets you want to work on, they automatically create worktrees and agents do the planning/research step, presenting you with plans like /grill-me so you can decide what to work on, alongside having a step for merging all of those worktrees whenever you want when a wave of work is done.
I haven't tried most of them tbh cause my OS install doesn't need to be even more of a dumpster, but I'm surprised at how none of these really seem all that... rock solid?
Like if I open a JetBrains IDE like IntelliJ IDEA / Rider I know that despite the crappy resource usage and whatnot, they have the needs of writing, refactoring, debugging and building code covered. Most of the current harnesses seem to get some baseline of telling agents what to do, but still somehow end up being buggy (e.g. can't peek into OpenCode sub-agents and give THEM instructions last I checked, or pause or alter the prompt, despite Claude Code allowing message exchanges with sub-agents and such), and many others are just clones of what already exists.
At the same time the desktop / web apps shipped with harnesses like OpenCode (if there's one at all) are very clearly insufficient and a step below Claude Code Desktop and Codex (ChatGPT app) in usability (like that move to horizontal tabs like browsers have, come on how am I supposed to navigate when there's like 30 of those? or what about an auto mode switch within the UI), so obviously we do need something as an ADE.
Personally I wonder if half of these projects will die off in like 5 years, or whether they'll get more polish or differentiate themselves more, and which will try to push more for cloud-only stuff, like how RooCode essentially dumped their VSC plugins and went all in on the cloud.
My macOS is a dumpster, so I can tell you that all these products do slightly different things. It’s apples and oranges across the board.
Except instead of just apples and oranges, there are several dozen fruits involved, and it takes a month to fully taste and digest any of them enough to decide whether they’re worth eating. And by then, there might be an improved cultivar of one of the fruits you tried before but decided wasn’t good enough.
How's this different to https://paseo.sh/ or https://www.onorca.dev/ which seems to be more feature complete? I personally use and like Paseo. I get to connect to my LLM agents even on other machines, remotely via Tailscale and the mobile app UX is quite good.
Huh. I just got pitched something like this earlier this week by an agent as well.
I'm not the target market for this, but this one has the same limitation as the one that pitched me: Mac Silicon only. No Linux, no Windows, no Intel Mac either.
Sibling comment mentioned Herdr, and Herdr is on all those platforms and apparently has raised $6 Mil.
The other catch for all of these: do you integrate with my custom harness? What do you offer that I can't vibe code myself?
You are right. You can build your own, or you can pick one you like and tweak it as you please.
I think different products cover different needs. I think there's lots of options nowadays and I'm most of the options shine one way or the other, it just depends on the mental model of the user and their goal.
I started my own tool less than amonth ago [1]. I did my first Show HN a week after, I was really proud, I added lots of features that first week just to accomodate my own needs and I was not even focused on this project but I could already see how it really helped me remove a lot of fatigue on managing multiple projects and be capable of manage them all at my own pace. My tool, in particular, was created with a focus on iterating over projects, not agents, and in the human (you) having full control on what you do. Anyway, the thing is that the very same day I posted the Show HN, there was this other project that was also submitted some hours before [3]. This project had more traction and comments already and was the one that appeared on the first page of HN. I haven't tried it, but I'm sure it would be great as well.
Yesterday I had a call with another dev at my organization, and he was also doing his own thing (unaware of the one I did, and unaware of other options available already) and I could already see the overlap with different tools and I even pointed out other projects for him to gather for ideas. I think is fantastic that so many people are exploring this space.
The reality I'm observing, with the latest state-of-the-art models, is that they are amazingly capable models and that less and less I need to even review what they produce. I trust on them more and more. This has deep implications moving forward.
I think we'll see two trends moving forward:
- Devs will build their own tools their own way (not all, but some will)
- Devs will clone an open-source project, and then, instead of forking and pusing upstream, they will just adjust to their own needs and detour from the original project more often than before
Sure enough, some people will just take the convenience of a ready-made project, and not even bother to iterate over your own thing, but when you are one prompt away from fixing this big issue or this subtle detail that annoys you, you don't need a config param anymore, and you don't need to wait days or weeks for somebody to fix, if you can compile the project in your machine, you can just create your own version of the software that works the way you want. You are just one prompt (or 10 or 50) of creating or adjusting to your own preferences.
Industry-wise I think this feels sometimes scary, but I think it is amazing.
Why would I want the terminal coding agents inside of a GUI app? The idea is terrible.
The first point in the explanations below makes no sense, claiming my agents will "keep working" if I close my notebook's lid or disconnect the network.
I don't think we even have a great understanding of the market, the niches, and where it is going.
It's also evolving very aggressively constantly, sticky workflows get replaced dramatically whenever new UX or tooling brings a shift into how things can be done.
I'm very curious if anyone has good research or idea of this space mapped out somewhere, I am trying to build for it too but it's really foggy right now and the North Star moves often.
These sorts of management tools really speed up development by automating all the PR management, such as making new worktrees, etc. and then super speed up the multitasking between different parallel feature development.
Excited to see if this one works better for me, and to try all the new alternatives in these comments.
[0]: https://goose.ai
[1]:https://paseo.sh/
[2]: https://t3.codes/
[3] https://orchestrator.inc/
[4] https://gitkraken.com/kepler
[5] https://emdash.com/
[6] https://superset.sh/
[7] https://www.conductor.build/
[8] https://nimbalyst.com/
[9] https://parallelcode.app/
[10] https://www.jetbrains.com/air/
There's also sometimes options that try to have a tighter integration with a single harness, which honestly seems to be better cause ACP is fairly bare:
[11] https://openchamber.dev/
Personally, I found Paseo to be very pleasant, except sometimes I had sub-agents kinda hang.
Kepler had a really nice issue tracking integration (GitLab/GitHub Issues, Trello, Linear, Jira) but their UI felt like vibe coded slop, like the core experience doesn't feel very good, but they keep adding shit that nobody needs like a graph visualization of what the agents are doing, instead of letting me drag the tasks around the Kanban board myself or tell the agents how to do that and how to transition the issue statuses. Hopefully it's gonna get better within a few years, still early on in development. Still, the idea of an issue tracker integration is brilliant and all harnesses should consider supporting that, e.g. pick which tickets you want to work on, they automatically create worktrees and agents do the planning/research step, presenting you with plans like /grill-me so you can decide what to work on, alongside having a step for merging all of those worktrees whenever you want when a wave of work is done.
I haven't tried most of them tbh cause my OS install doesn't need to be even more of a dumpster, but I'm surprised at how none of these really seem all that... rock solid?
Like if I open a JetBrains IDE like IntelliJ IDEA / Rider I know that despite the crappy resource usage and whatnot, they have the needs of writing, refactoring, debugging and building code covered. Most of the current harnesses seem to get some baseline of telling agents what to do, but still somehow end up being buggy (e.g. can't peek into OpenCode sub-agents and give THEM instructions last I checked, or pause or alter the prompt, despite Claude Code allowing message exchanges with sub-agents and such), and many others are just clones of what already exists.
At the same time the desktop / web apps shipped with harnesses like OpenCode (if there's one at all) are very clearly insufficient and a step below Claude Code Desktop and Codex (ChatGPT app) in usability (like that move to horizontal tabs like browsers have, come on how am I supposed to navigate when there's like 30 of those? or what about an auto mode switch within the UI), so obviously we do need something as an ADE.
Personally I wonder if half of these projects will die off in like 5 years, or whether they'll get more polish or differentiate themselves more, and which will try to push more for cloud-only stuff, like how RooCode essentially dumped their VSC plugins and went all in on the cloud.
Except instead of just apples and oranges, there are several dozen fruits involved, and it takes a month to fully taste and digest any of them enough to decide whether they’re worth eating. And by then, there might be an improved cultivar of one of the fruits you tried before but decided wasn’t good enough.
However, as many other comments have mentioned -- how do you plan to get users in a crowded space?
I'm not the target market for this, but this one has the same limitation as the one that pitched me: Mac Silicon only. No Linux, no Windows, no Intel Mac either.
Sibling comment mentioned Herdr, and Herdr is on all those platforms and apparently has raised $6 Mil.
The other catch for all of these: do you integrate with my custom harness? What do you offer that I can't vibe code myself?
I think different products cover different needs. I think there's lots of options nowadays and I'm most of the options shine one way or the other, it just depends on the mental model of the user and their goal.
I started my own tool less than amonth ago [1]. I did my first Show HN a week after, I was really proud, I added lots of features that first week just to accomodate my own needs and I was not even focused on this project but I could already see how it really helped me remove a lot of fatigue on managing multiple projects and be capable of manage them all at my own pace. My tool, in particular, was created with a focus on iterating over projects, not agents, and in the human (you) having full control on what you do. Anyway, the thing is that the very same day I posted the Show HN, there was this other project that was also submitted some hours before [3]. This project had more traction and comments already and was the one that appeared on the first page of HN. I haven't tried it, but I'm sure it would be great as well.
Yesterday I had a call with another dev at my organization, and he was also doing his own thing (unaware of the one I did, and unaware of other options available already) and I could already see the overlap with different tools and I even pointed out other projects for him to gather for ideas. I think is fantastic that so many people are exploring this space.
The reality I'm observing, with the latest state-of-the-art models, is that they are amazingly capable models and that less and less I need to even review what they produce. I trust on them more and more. This has deep implications moving forward.
I think we'll see two trends moving forward:
- Devs will build their own tools their own way (not all, but some will)
- Devs will clone an open-source project, and then, instead of forking and pusing upstream, they will just adjust to their own needs and detour from the original project more often than before
Sure enough, some people will just take the convenience of a ready-made project, and not even bother to iterate over your own thing, but when you are one prompt away from fixing this big issue or this subtle detail that annoys you, you don't need a config param anymore, and you don't need to wait days or weeks for somebody to fix, if you can compile the project in your machine, you can just create your own version of the software that works the way you want. You are just one prompt (or 10 or 50) of creating or adjusting to your own preferences.
Industry-wise I think this feels sometimes scary, but I think it is amazing.
[1] https://github.com/pausan/agenttik
[2] My Show HN 2 weeks ago: https://news.ycombinator.com/item?id=49720222
[3] Another, more successful, Show HN same day as mine: https://news.ycombinator.com/item?id=49713894
edit: formatting issues and minor tweaks on wording
Why would I want the terminal coding agents inside of a GUI app? The idea is terrible.
The first point in the explanations below makes no sense, claiming my agents will "keep working" if I close my notebook's lid or disconnect the network.
This is obviously not the case.