Encrypt data in use
Protect sensitive data during computation inside trusted execution environments (TEEs).
Caution is backed by Y Combinator. Learn more.
Take entire classes of attacks off the table. Deploy sensitive software in minutes with a familiar Git workflow, on your infrastructureinfra or ours.Take entire classes of attacks off the table. Deploy in minutes with a familiar Git workflow, on your infra or ours.
From engineers trusted by industry leaders
Protect sensitive data during computation inside trusted execution environments (TEEs).
Verify production matches reviewed source and build inputs through reproducible builds.
Deploy to confidential computing in minutes with Git, without building your own platform.
From source to a verifiable enclave in minutes. Watch this video for a full platform walkthrough.
Run caution init to create a caution.hcl and other data required for the application.
Push with git push caution main. Caution builds a reproducible
enclave image and provisions the TEE.
Run caution verify to rebuild the image, compare hashes,
and confirm exactly what the enclave is running.
Trust the operator. Production stays a black box.
Trust the hardware. The software stack still stays opaque.
Don't trust. Verify every layer yourself, down to the compiler.
From engineers trusted by industry leaders
For sensitive workloads where integrity failures would be costly—from custody and signing to confidential AI.
Show regulators and customers that key management and transaction signing happen only in attested, source-verified environments.
Serve LLMs and other models in verifiable environments so users can confirm the exact models and runtime that handled their prompts and data.
Give delegators and protocols independent proof that staking infrastructure is not forked, modified, or tampered with.
Run oracle and pricing services in verifiable enclaves to demonstrate the integrity of each price feed, signal, and the code that produces it.
Expose cryptographic proof that your full node binary matches the reviewed source and configuration.
Ensure training runs execute approved code on approved data, with attestation evidence for audit and compliance.
Prove that patient data never leaves a verified enclave, meeting HIPAA requirements with cryptographic evidence.
Caution runs across AWS Nitro today, with Intel TDX, AMD SEV-SNP, and TPM 2.0 attestations coming in 2026.
Caution is a secure hosting platform built on verifiable compute. You deploy applications through a familiar developer workflow, while Caution handles infrastructure setup, enclave builds, and deployment.
It protects data in use by isolating workloads from the surrounding host. Anyone with access to the source code and build inputs can also independently verify what is running in production.
If an attacker modifies your software during building or deployment, you can detect the change and block the modified software from receiving protected data or secrets.
The platform is fully open source.
Caution is suited to software that handles sensitive data or critical business logic, where a compromise would have serious consequences. Examples include:
Verifiable compute is a way to run software with cryptographic evidence that it matches reviewed source code, build inputs, and configuration.
It builds on confidential computing, where workloads run in isolated environments called secure enclaves or trusted execution environments (TEEs) that protect data while it is being processed. But isolation alone does not tell you whether the application inside is the one you intended to run.
Verifiable compute lets you verify this cryptographically by combining reproducible builds with hardware attestation. Anyone with access to the source and build inputs can rebuild the software and compare its measurements with signed evidence from the running enclave. This gives you cryptographic evidence that the deployed software matches the code you reviewed, helping you detect changes an attacker may have introduced during building or deployment.
Learn how verifiability and reproducible builds connect source code to a running deployment.
No. You can deploy an existing containerized application without rewriting it for Caution. Some changes may be needed to its build, deployment configuration, or how clients connect.
Independent verification requires a reproducible build. Your chosen encryption setup determines whether client integration is needed.
Our AI coding agent skills can help prepare your application, configure deployments, and troubleshoot issues.
Caution offers bring your own compute (BYOC) and fully managed hosting. Because Caution is open source, teams can also self-host it to operate deployments independently.
Compare the options in the deployment models docs.
Book 20 minutes with one of our engineers. We’ll discuss your goals, answer technical questions, and help you get ready to deploy with Caution.
Deploy confidential compute in minutes. Prove exactly what is running in production, from source code to live enclave.