Posted by karagenit 12 hours ago
Example: claude desktop + projects gives free compute for compilation without any cap (that I've found). So that + opus 5.5 which practically never runs out, and it's possible to get thousands of euros worth per month. On Pi/similar, the tokens + compute are metered and it quickly stacks up to something substantially more expensive.
Now this is clearly unsustainable, tip is to milk the cow for as long as it's possible
https://support.claude.com/en/articles/15036540-use-the-clau...
For those using Claude, there's pi-claude-code[0] which is currently still the way to use Ant subs in Pi.
https://support.claude.com/en/articles/15036540-use-the-clau...
Update October 7, 2026: Claude Max, Ultra, and Team plans now include monthly API credits. They cover the Claude API, Claude Managed Agents, and the Claude Agent SDK and claude -p when you run them with an API key from your linked Claude Console organization. Nothing changes for your plan: when you’re signed in with your Claude plan, the Claude Agent SDK and claude -p still draw from your subscription limits.
If you’ve used Codex before, then the Agent SDK is Anthropic’s equivalent of the Codex App Server, which is Codex running headlessly with a protocol to talk to it.
On My Pi is what I use for other models too: codex, opencode go, etc.
1) vibecoded, “works for me” level of stability: https://github.com/randomcodemonkey/autere
Not sure I'll actually switch but I'm curious. I've thought about trying to build my own harness and use Codex usage to work through it without spending a fortune on credits.
But OpenAI's subscription has seemingly always been available to use in other harnesses without trouble.
Pi/opencode apparently make 10% already of all traffic already:
https://manifest.build/blog/chatgpt-plus-tokens-third-party-...
All agents are ephemeral and running remotely now, and my phone has become my primary device to engage with agents.
(I'm using Ampcode and a variant of DeepSeek harness)
I gave the harness a try and it is refreshing, but the insistence on staying bare bones and not implementing basic functionality natively, together with no vetting process for popular 3rd party modules, made me return to Claude Code.
Wrap your agent of choice in e.g. docker sandbox, nono.sh, Nvidia OpenShell, microsanbox or any other sandboxing tool and run it without any checks (i.e. "yolo mode", dangerously-skip-permissions etc.)
IMHO it makes the whole category of problems around permission fatigue, trusting the harness (or plugin) authors etc go away.
Strong sandbox + disabling permission checks gives me maximum capabilities, least annoyance and (relative) peace of mind that any prompt injection, malicious plugin,... won't wreck my shit or leak credentials of any kind
I built a policy authorizer on top of pi-permission-system that looks kinda like nftables that models commands/subcommands/options as read/write and enfores policies depending on if the sub agent is read only or write, which really cuts down on the stupid decisions and nicely complements AI cmd review. Pi shines for making things like that possible.
*bare