--- title: "Caveman pricing" description: "Start saving for free. Go further with Caveman." canonical: https://caveman.so/pricing last-updated: 2026-09-30 --- # Caveman pricing Start saving for free. Go further with Caveman. Free local tools for smaller agent context. Model routing when you need it. Caveman Cloud to measure, evaluate, and improve your whole stack. ## Free local tools: $0 Fewer tokens. Already free. | Published result | Source | | --- | --- | | 33.2% fewer input tokens | [Proxy + Skill study](https://github.com/JuliusBrussee/caveman/blob/main/docs/WRAP-BENCHMARK.md) | | 63.6% fewer output tokens | [Elasticsearch Labs](https://www.elastic.co/search-labs/blog/elastic-caveman-ai-token-reduction) | Proxy + Skill benchmark: six Claude Code tasks, three paired runs each. Both groups passed 18/18 exact-answer checks. Input-token reduction, not a promised reduction in your bill. Elasticsearch Labs measured elastic-caveman across eight live MCP scenarios: 1,284 normal output tokens versus 467 Caveman output tokens. This is a separate study and workload from the Proxy + Skill input benchmark. - Without Caveman: 885,793 provider-reported input tokens. - With Proxy + Skill: 591,673 provider-reported input tokens. - [Read the benchmark](https://github.com/JuliusBrussee/caveman/blob/main/docs/WRAP-BENCHMARK.md). Free and open source: Apache-2.0 + MIT. No Caveman account or credit card required. Model-provider charges and your own infrastructure costs still apply. ### Caveman Skill Shorter answers. Cut filler from your agent’s replies. Keep code, commands, and errors intact. License: MIT. [Get the Skill](https://caveman.so/products/caveman) ### Caveman Proxy Smaller context. Compress what your coding agent reads. Keep the originals recoverable on your machine. License: Apache-2.0. [Get the Proxy](https://caveman.so/products/caveman-proxy) ### Caveman Middleware (Alpha) Your app, fewer tokens. Add local tool-result compression to your TypeScript or Python agent. Keep your framework. License: MIT clients · Apache-2.0 runtime. [Explore Middleware](https://caveman.so/products/caveman-middleware) ### Caveman Browse Lean browser tools. Give your agent a compact view of the browser, with original page details within reach. License: Apache-2.0. [Get Browse](https://caveman.so/products/caveman-browse) ## Community support Developers at Google, NVIDIA, Meta, Apple, GitHub, and Netflix starred Caveman. [Meet the community](https://github.com/JuliusBrussee/caveman/stargazers) ## Model Router Already sending less context? Explore which tasks can run on a more efficient model, while keeping demanding work on the models it needs. - Compare models against your tasks and quality bar. - Account for context, cache, cost, and latency. - Test routing changes before rolling them out. For developers and teams ready to optimize model choice. In private development. Discuss your workload, access, and pricing with the founder. No self-service paid checkout today. [Book with Julius](https://cal.com/caveman/chat) [Explore the Router](https://caveman.so/router) ## Caveman Cloud See what your agents cost, understand where they struggle, and test improvements before they reach production. One place for your team to turn evidence into better agents. - Trace runs, retries, failures, and spend across your stack. - Investigate problems and evaluate proposed fixes. - Manage routing, budgets, and controlled rollouts. For teams improving cost, quality, and reliability together. In private development. Discuss your workload, access, and pricing with the founder. No self-service paid checkout today. [Book with Julius](https://cal.com/caveman/chat) [Explore Caveman Cloud](https://caveman.so/products/platform) ### For enterprise Deployment options to discuss: Cloud-hosted, Private VPC, On-premises. Discuss deployment, SSO, and data retention with Julius. Scope the setup around your team’s requirements. [Enterprise details](https://caveman.so/enterprise) ## Questions ### What does free mean? The Skill, local Proxy, Middleware, and Browse have no Caveman subscription fee for your own use. Local compression needs no account or credit card. You still pay your model provider for inference and cover your own infrastructure. ### How are the free tools licensed? The local tools are free and open source. The Skill and middleware clients are MIT. The Proxy, Browse, and compression runtime are Apache-2.0. Check each repository for its full license terms. ### Will my bill drop by these percentages? These are separate studies: Caveman measured input tokens with Proxy + Skill; Elasticsearch Labs measured output tokens with elastic-caveman. The percentages do not add together. Results vary by workload, and cache pricing, retries, and recovery calls affect the final bill. ### How much do Router and Cloud cost? Both are in private development. Book with Julius, our founder, to discuss your workload, access, and pricing. There is no self-service paid checkout today, and using the free local tools does not require either product. ### Do I have to use every layer? Start wherever you need help. Use the Skill on its own, add the Proxy to your coding agent, or put Middleware in your application. Discuss routing or Cloud when model choice or team-wide visibility becomes your next problem. [Explore every product](https://caveman.so/products)