Support for Claude Code

[ajeetdsouza]

Hi, congrats on the launch! I know Claude Code support is being actively worked on, but I couldn't find any issue that was tracking it, so I created this. Thanks!

[jasperaelvoet]

#179

[gabrielMalonso]

I told Claude Code to do it for me, and it worked. It took me all morning to solve some bugs, but it worked. The only downside to this is that I have to check every update if it has conflicts.

[nassimna]

I told Claude Code to do it for me, and it worked. It took me all morning to solve some bugs, but it worked. The only downside to this is that I have to check every update if it has conflicts.

It's not about making it run the dev has it working in other branch but the problem Anthropic has to approved else users could get banned same as what happend in opencode

[gabrielMalonso]

I told Claude Code to do it for me, and it worked. It took me all morning to solve some bugs, but it worked. The only downside to this is that I have to check every update if it has conflicts.

It's not about making it run the dev has it working in other branch but the problem Anthropic has to approved else users could get banned same as what happend in opencode

Nobody will get banned over this. I've been using conductor for a long time, and nothing ever happened. I know they are sketchy, but I don't believe they will ban people for using Claude Code in a wrapper. I could be wrong, tho.

[akarabach]

Using OpenCode Desktop for 4-5 weeks, everything is fine (except I need to authenticate every 10-12 hours), so, still not banned.

Would be nice to have it with BIG RED warnings message that we can be banned, it's fine ;)

[chuks-qua]

Using OpenCode Desktop for 4-5 weeks, everything is fine (except I need to authenticate every 10-12 hours), so, still not banned.

Would be nice to have it with BIG RED warnings message that we can be banned, it's fine ;)

You know Theo is a big UX guy, i know he won't be pleased with this solution

[akarabach]

Using OpenCode Desktop for 4-5 weeks, everything is fine (except I need to authenticate every 10-12 hours), so, still not banned.
Would be nice to have it with BIG RED warnings message that we can be banned, it's fine ;)

You know Theo is a big UX guy, i know he won't be pleased with this solution

Without Claude Code, the T3 Code is basically a no-go for ~60+++% of developers

[chuks-qua]

True, but the choices are: risk your app the cause of devs getting their accounts bad and tuple ur reputation or wait for Anthropic. the longer they take to respond the worse they look

[copypasteitworks]

My current understanding is that it depends on the implementation basically whether you're using:

• CC subscription OAuth token for your own requests/SDK, which most likly is against TOS and therefore risky

• CC CLI supported programatic use via headless mode

I also implemented quick and dirty CC support using the later, based on that i would assume the main challenge to be mostly architectural due to vastly different control surface exposed by cc vs. codex:

• codex supports create app servers directly (noice)
• cc limits programmatic use to headless cli invoke explicitly (https://code.claude.com/docs/en/headless) for subscriptions while the SDK (https://platform.claude.com/docs/en/agent-sdk/typescript) only via API (my assumption please tell me if not true!) leaving us basically with something like:

session_id=$(claude -p "fix all and make it super safe (nuclear option)" --output-format json | jq -r '.session_id') claude -p "I said no mistakes" --resume "$session_id"

@akarabach @gabrielMalonso @chuks-qua @nassimna