You don't have to pay more - I wouldn't switch to an open source harness on a whim, but your harness/cheaper agents can always call Claude Code headlessly.

You can use your claude subscription just fine in open harnesses!

I make one (https://github.com/juggler-ai/juggler) and it'll auto-detect and use your claude CLI. I use that claude plan for 90% of my work, plus a bit of GPT, I just prefer to do it inside a nicer, less locked-in environment than claude.

There's really no moat around proprietary closed harnesses, it's trivial to jump between them, but people need to be aware that it's possible!

Pi can utilize Claude or GPT subscriptions and they work wonderfully. You can have both subscriptions logged in and write your own tools to route to the model that works best in every scenario. I have access to both pro level subscriptions for work and a large credit count for api calls for Claude. I use all 3 with JEV as my router with a local model for context management and compaction. my token usage has gone down to a point where I rarely even use api calls. I would recommend oh-mi-pi for anyone who wants to get started with using an open source harness as it has very good memory and context management tools that ship with it and are extremely easy to set up.

>4. Although it's hard to 100% confirm this, there's reason to believe that Claude itself works better with Claude Code than with other harnesses.

I thought this was pretty much well established that all of the labs have model - harness pairings for exactly this reason

Whale Paper. https://arxiv.org/abs/2609.00196

EMNLP talk from like 2 weeks ago https://aclanthology.org/2026.findings-emnlp.2207/

You realize this is about Cursor, not Claude Code?

I created hlock https://docs.hlock.ai/ to exactly tackle some of these problems: it lets you port your harness configuration across the different harnesses (open source and proprietary)

yep, laughing so hard as all these thought-leaders are racing to make their companies essentially subsidiaries of anthropic/openai/etc because they are structuring their entire software around humans never looking at it...so basically they just hired anthropic chatbots to do everything.....when those prices go up or those companies go out of business there are going to be a lot of companies completely paralyzed despite all the fart sniffing and high horse riding they were doing. I think that is the whole point of the marketing of these fancy new models and shit....convince companies to get 100% addicted to it and perhaps even lose the in-house staff to work without it.....then whatever it costs they will have to pay.

They spend more time trying to figure out how to put more agents into their "process" and solve any problems it creates with more agents than just doing a general claude run and tweaking it themselves with the skills they built over their career ostensibly.

Cracks me up watching engineers completely degrade themselves by saying that a "team" of "agents" should just be let loose on a task and create 10,000+ line PRs that no human will ever read, then deploy it to production and run code no human has seen or understands....like, stop degrading yourself to prop up these companies bleeding money....it is a tool, it is powerful, it isn't an engineer though and you are, it isn't smart and you are, don't intentionally argue for your own replacement by a fucking corporate chatbot.

If I had to work at a company that thought they could do software that way I would refuse to join the on-call rotation....if that fucking shit is so much better than me that we need to rush un-reviewed shit to production then I'm sure it is also so much better than me and it can manage oncall for nights and weekends at least...unassisted. Watch how many of these zealots change their tune when you tell them they should hand over on-call 24/7 in production to bots and not even put people in a pager rotation or page them. Suddenly real people are needed to fix the software we were too stupid to write but the bot - that did write it - is too fucking stupid to manage?

make it make sense.

You can get massive productivity increases with these tools while still using manual-mode only, reading the code as it is generated via the diffs, yell at it when it does boneheaded shit (which it 100% will), and doing narrow-scoped tickets/PRs that are reviewable by a human being just like in the before times. What is the need to go faster than that and essentially just tie the entire product of your company to what some other company's bot chooses to shit out to you that day?

It's the one features that keeps me using claude, it's simply convenient.

You can. Anthropic documents it: the Agent SDK and claude -p work signed in with your Claude plan and draw from your normal subscription limits. https://support.claude.com/en/articles/15036540-use-the-clau...

We build a desktop app on top of the SDK, so I've spent a lot of time on that page.

You're wrong.

Only if you use Claude via /login in pi... in which case

  claude setup-token

;)

say more, what do you think CC does better?

OpenCode + DeepSeek V4.1 Flash Uncensored on Venice AI is truly incredible. Granted, I have it also drive an interactive claude code session in tmux and just do everything Claude Code refuses to. But, it just gets shit done. Kind of what makes Instinct so powerful, don't tell me you won't do something, just get it done.

What's your reasoning? What makes Claude Code better? Genuinely curious to learn.

What if I'm trying to bypass the security monitor built-into CC?

I wish you would explain your reasoning instead of preemptively telling people they don't know what they're talking about. Then I might actually learn something.