Top
Best
New

Posted by rbanffy 1 day ago

OpenCode – Open source AI coding agent(opencode.ai)
1182 points | 578 commentspage 4
justacatbot 9 hours ago|
The decision to build this as a TUI rather than a web app is interesting. Terminal-native tools tend to get out of the way and let you stay in flow -- curious how the context management works when you have a large codebase, do you chunk by file or do something smarter?
solarkraft 7 hours ago|
It’s both! The core is implemented as a server and any UI (the TUI being one) can connect to it.

It’s actually “dumber” than any of your suggestions - they just let the agent explore to build up context on its own. “ls” and “grep” are among the most used discovery tools. This works extraordinarily well and is pretty much the standard nowadays because it lets the agent be pretty smart about what context it pulls in.

rurban 9 hours ago||
That's my favorite CLI agent, over codex, claude, copilot and qwen-code.

It has beautified markdown output, much more subagents, and access to free models. Unlike claude and codex. Best is opencode with GitHub opus 4.6, but the fun only lasts for a day, then you're out of tokens for a month.

frankdejonge 14 hours ago||
I've used it but recently moved back to plain claude code. We use claude at the company and weirdly the experience has become less and less productive using opencode. I'm a bit sad about it as it was the first experience that really clicked and got great results out of. I'm actually curious if Anthropic knows which client is used and if they negatively influence the experience on purpose. It's very difficult to prove because nothing about this is exact science.
ec109685 6 hours ago|
I think Anthropic just highly RL’s their model to work best with it’s Claude Code’s particular ways of going about things.

All the background capability Claude code now has makes things way more complex and I saw a meaningful improvement with 4.6 versus 4.5, so imagine other harnesses will take time to catch up.

madduci 17 hours ago||
Have they "squatted" the name? It's the same name for the digital Sovereignty initiative in Germany

https://opencode.de/

embedding-shape 13 hours ago||
If it was a somewhat unique name, then yeah maybe. But "opencode" is probably as generic as you could make it, hard to claim to be "squatting" something so well used already... Earliest project on GitHub named "opencode" seems to date back to 2010, but I'm sure there are even earlier projects too: https://github.com/search?q=opencode&type=repositories&s=upd...
heywinit 17 hours ago||
you'll be surprised the name was actually a controversy on x/twitter since opencode was originally another dev's idea who joined the charmcli team. they wanted to keep that name but dax somehow (?) ended up squatting it. the charmcli team has renamed their tool to "crush" which matches their other tools a lot better than "opencode"
dominotw 12 hours ago||
oh yea that whole drama turned me off from this project. dax guy seems to be some sort of grumpy cat.
boomskats 14 hours ago||
I've been using opencode for a few months and really like it, both from a UX and a results perspective.

It started getting increasingly flaky with Anthropic's API recently, so I switched back to Claude Code for a couple of days. Oh my, what a night and day difference. Tokens, MCP use, everything.

For anyone reading at OpenAI, your support for OpenCode is the reason I now pay you 200 bucks a month instead.

embedding-shape 14 hours ago|
I've been paying OpenAI 200 bucks a month for what feels like forever by now, but used OpenCode for the first time yesterday, been using Codex (and Claude Code from time to time, to see if they've caught up with Codex) since then.

But I don't use MCP, don't need anything complicated, and not sure what OpenCode actually offers on top. The UI is slightly nicer (but oh so much heavier resource usage), both projects source code seems vibecoded and the architecture is held together with hopes and dreams, but in reality, minor difference really.

Also, didn't find a way in OpenCode to do the "Fast Mode" that Codex has available, is that just not possible or am I missing some setting? Not Codex-Spark but the mode that toggles faster inference.

Fabricio20 20 hours ago||
I wish they would add back support for anthropic max/pro plans via calling the claude cli in -p mode. As I understand thats still very much allowed usage of claude code cli (as you are still using claude cli as it was intended anyway and fixes the issue of cache hits which I believe were the primary reason anthropic sent them the c&d). I love the UX from OpenCode (I loved setting it up in web mode on my home server and code from the web browser vs doing claude code over ssh) but until I can use my pro/max subscription I can't go back, the API pricing is way too much for my third world country wallet.
griffiths 16 hours ago||
They had that?! I saw that some people wrote skills and plugins to call claude cli and gemini cli to still be able to use the subscription. I would also wish that this was supported out of the box, something similar to goose cli providers or acp providers (https://block.github.io/goose/docs/guides/acp-providers). But I don't want to spend testing yet another agent harness or change the workflow when I somewhat got used to one way of working on things (the churn is real).
unixfox 15 hours ago||
I guess you could look into my plugin for that use case of CC inside opencode: https://github.com/unixfox/opencode-claude-code-plugin
zingar 1 day ago||
Anecdotal pros and one annoyance:

- GH copilot API is a first class citizen with access to multiple providers’ models at a very good price with a pro plan - no terminal flicker - it seems really good with subagents - I can’t see any terminal history inside my emacs vterm :(

tomasz-tomczyk 11 hours ago||
I'd love for all these tools to standardise on the structure of plugins / skills / commands / hooks etc., so I can swap between them to compare without feeling handicapped!
lordforever7 3 hours ago||
doesn't someone feel weird with the tui being not that of terminal but more of UI with text box coming in center, etc?
hereme888 1 day ago|
The reason I'm switching again next month, from Claude back to OpenAI.
hungryhobbit 1 day ago||
Yeah, support the company that promised to help your government illegally mass surveil and mass kill people, because they support a use case slightly better than the non-mass-murdering option.
stavros 1 day ago|||
Both of them promised to help their government illegally mass surveil and mass kill people. One of them just didn't want it done to US citizens.

I'm not a US citizen, so both companies are the same, as far as I'm concerned.

hungryhobbit 1 day ago||
You are absolutely correct that both are evil ... as are most corporations.

Still, I feel like "will commit illegal mass murder against their own citizens" is a significant enough degree more evil. I think lots of corporations will help their government murder citizens of other countries, but very few would go so far as to agree to murder their own (fellow) citizens ... just to get a juicy contract.

stavros 1 day ago||
I see your viewpoint but, to me, "both will happily murder you but one is better because they won't murder ME!" isn't very compelling. Like, I get it, but also it changes nothing for me. They're both bad.
hungryhobbit 1 day ago|||
It's not about "won't murder me" it's about "won't murder their own tribe". Humans are very tribal creatures, and we have all sorts of built-in societal taboos about betraying our tribe.

We also have taboos against betraying/murdering/whatever people of other tribes, but those taboos are much weaker and get relaxed sometimes (eg. in war). My point is, it takes significantly more anti-social (ie. evil) behavior to betray your own tribe, in the deepest way possible, than it does to do horrible things to other tribes.

This is just as much true for Russians murdering Ukranians as Ukranians murdering Russians, or any other conflict group: almost all Russians would consider a Russian who helps kill Russians to be more evil than a Russian who kills Ukranians (and vice versa).

stavros 1 day ago||
Right, but I consider someone who'll murder exclusively other tribes to be infinitely closer to someone who'll murder their own tribe than to someone who won't murder anyone.
cyanydeez 1 day ago|||
watching trump get elected twice; you can see why americanos have no problemos with mental backflips when choosing.

But you're still choosing evil when you could try local models

kykat 1 day ago||
Will you send me an H100?
cyanydeez 1 day ago||
Are you doing something that actually demands it? Have you tried local models on either the mac or AMD395+?
kykat 1 day ago|||
I will be able to do something that demands it once I have it ;)
cyanydeez 11 hours ago||
Most people who win the lottery are poor again within the decade.
HWR_14 1 day ago|||
Will you send me an AMD395+ or a new Mac that can handle the local models? That would probably be enough for me.
hereme888 8 hours ago||||
That a gross exaggeration. But to your point, I could say the same for almost any product I use from Big Tech, every laptop company I buy my hardware from, etc. I'm sure the same applies to you. I can't fight every vendor all the time. For now I pick what works best for my use case.
Robdel12 1 day ago|||
> mass kill people

https://www.washingtonpost.com/technology/2026/03/04/anthrop...

xvector 23 hours ago||
You're right, Anthropic shouldn't have even taken a moral stance here at all. They should have just gone full send and allowed everything, because there will never be satisfying some people. Why even try?
More comments...