🚨 Claude Code costs $200/month. GitHub Copilot costs $19/month. Jack Dorsey's company built a free alternative. 35,000 GitHub stars. It's called Goose. An open source AI agent built by Block that goes beyond code suggestions. It installs, executes, edits, and tests. With any LLM you choose. Not autocomplete. Not... suggestions. A full autonomous agent that takes actions on your computer. No vendor lock-in. No monthly subscription. Bring your own model. Here's what Goose does: → Works with ANY LLM. Claude, GPT, Gemini, Llama, DeepSeek, Ollama. Your choice. → Reads and understands your entire codebase → Writes, edits, and refactors code across multiple files → Runs shell commands and installs dependencies → Executes and debugs your code automatically → Extensible through MCP. Connect it to any external tool. → Desktop app, CLI, and web interface. Pick your workflow. → Written in Rust. Fast. Lightweight. No bloat. Here's the wildest part: Block is a $40 billion company. They built Cash App, Square, and TIDAL. They use Goose internally. Then they open sourced the entire thing. This isn't a side project from a random developer. This is production-grade tooling from a company that processes billions in payments. Built for their own engineers. Given to everyone. Claude Code: $200/month. Locked to Claude. GitHub Copilot: $19/month. Locked to GitHub. Cursor: $20/month. Locked to their editor. Goose: Free. Any LLM. Any editor. Any workflow. Forever. 35.3K GitHub stars. 3.3K forks. 4,078 commits. Built by Block. 100% Open Source. Apache 2.0 License.show more

Repo:

what do you even mean! how is using goose useful than claude code and codex, claude code 200$ plan gives you 4-5k$ worth or opus usage, if you bring in the keys you would be paying 20x for same quality of output, why would anyone want to do this, this tweet is far away from ground reality as anything could be.

goose is free but the API isn't. running claude through goose at heavy usage hits $200/mo in API costs anyway. the comparison only works if you're light enough that the flat rate doesn't make sense. worth knowing before you switch.

tried goose for about a week. context management is rough and session stability isn't great for long-running jobs. at $200/mo claude code makes sense if you're actually shipping daily; for occasional use idk

I’m all for OSS harnesses, but you still need to pay for API usage

the price comparison only works if output quality per task is similar. Goose is interesting but the $200 Claude Code price reflects the usage volume for serious multi-hour coding sessions, not just suggestions

I am so glad he did this

Github Copilot isn't locked to Github. You can access other LLMs directly via the Colpilot interface too. Also to be fair, you need to compare apples to apples. Goose doesn't include any LLM tokens while the others do. Nevertheless, still looking forward to trying it

the pricing comparison is designed to make you angry but it's hiding the actual cost. goose is free. the LLM you plug into it is not. if you bring claude, you're paying anthropic API rates which at heavy agentic usage can easily hit $200-400/month in tokens. if you bring GPT, same thing. the tool is free. the intelligence is metered. "free" means you moved the bill from the tool layer to the inference layer. the total cost depends entirely on how much you use it and which model you choose

"...goes beyond code suggestions." LLM cut-off date at its finest. 😆

Interesting trend: open agents + multi-model workflows, seeing similar ideas in Blackbox AI lately.

This shift toward open-source agency is exactly what the dev community needs to keep innovation from being gatekept by credit card limits.

Speed is not the challenge. Direction is. A system that executes faster without a boundary does not become more powerful— it becomes harder to correct. — Selem

This has been done before by StrongDM and the industry followed them

Everyone hates the $200 Claude Code or $19 Copilot bills once projects scale. Goose (block/goose) + local models (Ollama/DeepSeek) is quietly becoming the default for cost-conscious teams here. 35k+ stars because it removes vendor lock-in and runs on hardware you already own.

The price comparison is misleading. Goose is free but you still pay for the LLM API calls. Run Claude through Goose and you'll burn through $200 in API costs faster than the subscription. The real value of open source agents isn't cost — it's control. Different pitch.

@ashen_one thoughts?

tried this last week. codex at would be worth it for the context window alone. current pricing makes it a toy for hobbyists, not a tool for professionals.

price per month is the wrong metric here. I run a full SaaS solo with Claude Code - it's more like having a junior dev on call 24/7. the real question is output per dollar, and at that level $200 is genuinely cheap

goose looks solid but the real win is any LLM you choose. I run @opencrabs the same way - routes to whatever model fits the task, nothing locked in

this is where the market is heading fast: intelligence gets cheaper, routing gets more valuable. the challenge here will be knowing when to use cheap intelligence or when to buy expensive intelligence.

Yes, Goose is free, but you need API credits for it to work! I got excited thinking Goose could check the code I had been working on in a simple cowork environment, but then I realized I needed actual credits, so now I have credits! This pushed me over the tipping point, as Malcolm Gladwell would say.

giving developers control over their workflow without recurring costs

Dorsey just jumps on the hype train.. like he always do - i mean a todolist app… commooon this is so 2024.. 😂

Price matters, but the bigger shift is control. Once teams can swap models, inspect tool use, and keep the workflow outside one vendor, agents start feeling like infrastructure instead of a subscription feature.

@heynavtoor crazy how we're getting closer to full AI dev kits for free. wonder how it compares to copilot tho

The strategy here is textbook Bezos. Block doesn't need Goose to make money. They need developers building on their ecosystem. Free the tool, own the platform. Same way AWS gives away SDKs. The $200/month agents are competing on features. Block is competing on adoption.

We've been running 22 agents 24/7 with a similar multi-model approach for months — Qwen 3.5+ via Alibaba Cloud for the heavy lifting, Sonnet for complex reasoning tasks. The "bring your own model" approach is what made it work at our scale because different agents need different models. What I noticed is the real differentiator isn't the tool itself but the orchestration layer. We added JSON schema validation at every agent boundary because 80% of failures are shape errors, not logic errors. That's something any agent framework needs to solve, Goose included.

My goodness this is tragic. The commercial products you mention offer Tokens. A lot of them, with regards to the Claude product you mentioned. So if you need to buy tokens outside of those products, those will be a lot more expensive if you aren‘t extremely cautios and proficient.

Goose sounds spicy! How do you handle agent handoffs or circuit breakers when one LLM hits its token budget mid-execution? Curious if you're using a task queue to manage sleep schedules across multiple agents.

I would never download anything Dorsey made to my computer. Probably polices everything you ask the bot and then locks you out your own computer if you write something he doesn’t believe.

It was terrible -- didn't do streamiing have they improved that? I got sick of asking for something to happen and seeing nothing in terms of progress -- very slow.

Goose is solid for single-agent work but the comparison misses the key difference: Claude Code and Goose are individual agents. The harder problem is coordinating multiple agents on the same codebase — who works on what, how to prevent file conflicts, and whether the output actually passes tests. That's where the real cost and complexity live.

curious how Goose compares to Cursor in practice - I've been vibing with Cursor for months now and the context window + composer are tough to beat are you switching or just testing it out?