SUPERCODE BETA IS LIVE. 🚀
https://t.co/D8yAspfmNQ
An open-source AI agent in your terminal that controls your whole machine.
In beta now:
• 5 free models included
• bring your own API key
• action commands
• full OS access — read files, run commands, edit code, open apps, search the web
• persistent memory across sessions
• you approve every action on your command.
Coming in next beta versions:
• screen-actuation agent (sees your screen, hears your voice, moves your cursor)
• MCPs & more providers
• app integrations
• and more
try it: npm install -g supercode-cli@latest
Hey team @ollama 👋
We at @supercodeai are building an open-source harness agent with 2380+ users.
We just crossed 110M monthly token usage, 21K+ monthly downloads and 163+ Github stars — and the demand for open models is accelerating more.
We want to make it dead-simple for for our users to run open models inside Supercode through Ollama.
looking to collaborate and work with you to host more open models on Supercode so users can harness them natively.
cc: @jmorgan@mchiang0610@devonrifkin@parthsareen
@zachmoskow@pinecone@concentrateai If multiple organizations join for hosting an irl event then maybe we can get more than 100-150(depending upon the kind of event)
@george_onx@opencode hey George, we're building @supercodeai - an open source harness agent with over 2700 active users, and we're want to work with Pioneer team.
discuss this over DMs!?
Almost missed this: @daytonaio has signed the open letter supporting open-weight AI models
One of the reasons the US leads in AI is because builders here share and collaborate.
The Open Weights and American AI Leadership declaration is a testament to that: https://t.co/WUfVKpf05M
Introducing Supercode Crisp 🚨
when you ask an AI to add a config system, it gives you a factory pattern, an interface, and a parser — for a value that never changes. The code works. But it's 10× more than you needed, and you're paying for every token.
This is where Crisp comes in-
Supercode Crisp is a ruleset that makes your AI write the least code that works — like a senior dev who's been paged at 3am one too many times.
Improvements:
→ 6–20% of original code size
→ 77% fewer dependencies
→ 3–6× faster builds
→ Fewer tokens per task
read blog for more details: https://t.co/39KcekMuve
We launched BYOK on Concentrate. And it's FREE.
At $10/M tokens, 10M requests / month, and 2K tokens/request, @OpenRouter charges you $9K.
Concentrate is $0. Thanks to our YC customers for this request!
https://t.co/AcI0yemNJW