LeanStack
LeanStack is a team or company focused on agent engineering practice. It publicly shares its cost architecture and engineering experience for large-scale agent requests. Its core approach is to place a local gateway between agents and model providers for routing management.
Timeline (last 90 days)
Between September 15 and October 15, 2026, LeanStack's 66 scheduled Agent tasks generated 276,814 model requests, consuming 9.9 billion input tokens and 385 million output tokens, of which 7.2 billion input tokens came from cache hits (about 73%), with a total bill of 430 US dollars in fixed subscription fees and zero token-based API spending.
LeanStack places a local gateway between the Agent and all model providers, and all Agents, scripts and coding tools communicate only with this single local endpoint, with the gateway deciding where requests actually go.
The gateway architecture brings four benefits: all requests are written to the same log (including provider, model, status code, latency and token usage); routing policies are centralized in one place, and switching from a paid to a free route requires changing only one line of configuration; when a route fails, alternatives can be tried automatically; and multiple routes run on fixed-rate subscription logins rather than pay-as-you-go API keys, which is the source of the zero cost per token.
LeanStack's monthly fixed spending is: Claude plan 200 US dollars, Cursor plan 200 US dollars, ChatGPT plan 20 US dollars, coding plan 10 US dollars, totaling 430 US dollars; token-based API spending is zero.
Of LeanStack's 276,814 requests, 71,953 did not return successfully, including 22,111 HTTP 429 rate limits, with the rest including 404s caused by model renaming or retirement, upstream 5xx errors, request errors and client disconnections; some free routes have a success rate far below half.
When monthly usage is only a few hundred calls, pay-as-you-go is almost certainly cheaper than the 430-dollar plan.
The LeanStack team is organizing this approach into a 29-dollar playbook, which is currently still in the demand-testing stage.
All Facts (3)
LeanStack places a local gateway between the Agent and all model providers, and all Agents, scripts and coding tools communicate only with this single local endpoint, with the gateway deciding where requests actually go.
50%UnverifiedThe gateway architecture brings four benefits: all requests are written to the same log (including provider, model, status code, latency and token usage); routing policies are centralized in one place, and switching from a paid to a free route requires changing only one line of configuration; when a route fails, alternatives can be tried automatically; and multiple routes run on fixed-rate subscription logins rather than pay-as-you-go API keys, which is the source of the zero cost per token.
50%UnverifiedWhen monthly usage is only a few hundred calls, pay-as-you-go is almost certainly cheaper than the 430-dollar plan.
50%