vje013/darwin-agentic-cloud

Darwin Agentic Cloud is a platform that lets AI agents run code and take actions in the cloud, with every step cryptographically signed and verifiable. The basic idea is that when an AI agent does something on your behalf, such as executing a script or making a purchase, you currently have no way to prove what actually happened. You have to trust that the framework did what it said. Darwin generates a signed receipt for each execution that anyone can independently verify, which makes the agent's behavior auditable rather than just trusted. The platform is designed for businesses that want to let AI agents make purchases or take automated actions without losing control over spending. You authorize a cost cap, and the agent can only spend up to that limit. If a website or service is not set up to work with agents reliably, Darwin does not send the agent there. The project claims it is free to users and earns from merchants rather than from the people running agents. From a developer's perspective, the API is a single Python function. You call it with the code you want the agent to run, and it returns an attestation object containing the output, the cost, and a signed key identifier. A command-line tool can then verify that attestation against a public key list published by the project. The receipt includes details like which hardware substrate the code ran on, how long it took, and whether it stayed within the authorized cost cap. The project is in a sandboxed beta and research phase starting June 2026. It requires Python 3.11 or newer and can be installed from PyPI with a single pip command. Sign-in uses a Google account, and no new API keys are required. The repository is open source under the Apache 2.0 license.