Claude Code's $200 Monthly Bill Meets a Free Rival: Goose

Claude Code's $20-to-$200 monthly price tag has developers hunting for a free alternative. Block's open-source Goose agent wants to be it.

What changed

Claude Code, Anthropic's terminal-based agent that writes, debugs, and deploys code autonomously, has captured developers' attention worldwide. Its pricing has not, sparking what a report describes as a growing rebellion among the programmers it serves.

Goose is Block's answer — the fintech company formerly known as Square. It offers close to the same functionality as Claude Code but runs entirely on a developer's own machine.

No subscription. No cloud dependency. No rate limit that resets every five hours.

“Your data stays with you, period,” said Parth Sareen, a software engineer who demonstrated Goose during a recent livestream. The line sums up the pitch: full control over an AI workflow, including the ability to work offline — even on a plane.

For engineering teams weighing the switch, that pitch is bigger than nostalgia for command-line tools — it's a promise that the workflow keeps working when the network, or the subscription, does not.

The project has momentum to back the pitch. Goose has passed 26,100 stars on GitHub, with 362 contributors and 102 releases since launch. Version 1.20.1 shipped January 19, 2026 — a release cadence that matches commercial software.

For engineering teams watching adoption curves, that release cadence is itself a signal — an actively maintained open project reduces the risk of betting on a tool that stalls.

Claude Code's pricing tiers

Claude Code's pricing, by contrast, tightens the more a developer actually uses it:

Plan	Price	Prompts per 5 hours	Notes
Pro	$17/mo (annual) or $20/mo	10–40	Entry tier
Max	$100/mo	50–200	Includes Claude 4.5 Opus access
Max	$200/mo	200–800	Includes Claude 4.5 Opus access

Serious developers exhaust the Pro tier's 10-to-40-prompt window within minutes of intensive work, per the same report.

Why the rate limits are a sore point

In late July, Anthropic layered on weekly rate limits, according to the report. Pro subscribers get 40 to 80 hours of Sonnet 4 usage per week; $200 Max users get 240 to 480 hours of Sonnet 4 plus 24 to 40 hours of Opus 4.

Independent analysis cited in the report translates those hours into roughly 44,000 tokens per session for Pro users and 220,000 tokens for the $200 Max plan. “It's confusing and vague,” one developer wrote in a widely shared analysis. “When they say ‘24-40 hours of Opus 4,’ that doesn't really tell you anything useful about what you're actually getting.”

Anthropic has defended the changes, saying they affect fewer than five percent of users and target people running Claude Code continuously in the background, 24/7.

Trying Goose locally

Goose is designed to pair with local models, so getting started does not require an Anthropic account. A typical flow — check Goose's own documentation for exact commands and current flags — looks like this:

1. Install a local model runner such as Ollama and pull a coding-capable model.
2. Install Goose using the project's installer script.
3. Run goose configure to point it at the local model endpoint.
4. Start a session in the project directory and hand it tasks the way you would with Claude Code.

# illustrative example — verify exact syntax in Goose's docs
ollama pull qwen2.5-coder
goose configure
goose session start

The trade-offs

Goose is not a drop-in upgrade. One developer who switched to Claude Code's $200 plan described the gap in output quality bluntly: “When I say ‘make this look modern,’ Opus knows what I mean. Other models give me Bootstrap circa 2015.”

That is the honest trade: local, model-agnostic tools like Goose hand developers control over cost and data, but the model doing the work is still whatever they connect it to. Anthropic's frontier models remain a specific advantage Claude Code sells directly, and swapping in an open model changes the ceiling on output quality, not just the price. Teams already invested in Claude Code's workflow tooling also give up integrations and support bundled with the subscription — a cost that never shows up on an invoice.

Part of a broader shift

Goose is not the only open alternative gaining ground. NousCoder-14B, a coding model released by Nous Research, arrived at what a second report calls a particularly charged moment for Claude Code, which has dominated social media discussion since New Year's Day.

Jaana Dogan, a principal engineer at Google responsible for the Gemini API, captured that mood in a viral post on X: “I gave Claude Code a description of the problem, it generated what we built last year in an hour.”

NousCoder-14B posts a 67.87 percent accuracy rate on LiveCodeBench v6, a benchmark built from competitive programming problems published between August 2024 and May 2025 — a 7.08 percentage point gain over its base model, Alibaba's Qwen3-14B, according to Nous Research's own technical report.

Whether any of these open projects replace Claude Code outright is a separate question from whether they change what developers expect to pay for one.

What to watch

• Whether Anthropic loosens or further tightens Claude Code's weekly rate limits after user pushback over their clarity.
• How fast Goose's GitHub stars and release cadence keep climbing past the 26,100-star, 102-release mark it has already hit.
• Whether open coding models close the quality gap that pushes developers back toward Claude 4.5 Opus for front-end and design-heavy tasks.