# OpenReward

## Docs

- [Architecture Overview](https://docs.openreward.ai/concepts/architecture.md): Understanding OpenReward environments and sandboxes
- [Environments](https://docs.openreward.ai/concepts/environments.md): Long-running ORS servers with automatic scaling
- [Sandboxes](https://docs.openreward.ai/concepts/sandboxes.md): Execution containers for agent code
- [GitHub Deployment](https://docs.openreward.ai/deployment/github-integration.md): Connect and deploy environments from GitHub repositories
- [Local Development](https://docs.openreward.ai/deployment/local-development.md): Test environments locally before deploying to OpenReward
- [Using the Async Client](https://docs.openreward.ai/environments/async-client.md): Learn to use the asynchronous features of OpenReward
- [Building Agentic Environments](https://docs.openreward.ai/environments/building-agentic-environments.md): Make and deploy your first agentic environment
- [Debugging Environments](https://docs.openreward.ai/environments/debugging-environments.md): Tips and techniques for debugging your environments
- [Deploying Harbor Environments](https://docs.openreward.ai/environments/deploying-harbor-environments.md): Deploy environments that use the Harbor task specification
- [Docker Images](https://docs.openreward.ai/environments/docker-images.md): Choose and configure Docker images for sandbox environments
- [Environment Cards](https://docs.openreward.ai/environments/environment-cards.md): Document your environment so users know what it does
- [Your first evaluation](https://docs.openreward.ai/environments/evaluation.md): Learn how to evaluate models with OpenReward
- [Making GPU Environments](https://docs.openreward.ai/environments/gpu-environments.md): Configure environments that use GPU-accelerated sandboxes
- [Keeping Secrets Secret](https://docs.openreward.ai/environments/keeping-secrets-secret.md): How OpenReward protects sensitive API keys using placeholder injection and egress-time substitution
- [Using Environment Variants](https://docs.openreward.ai/environments/using-environment-variants.md): Serve multiple environments from a single server and select them by variant
- [Using LLM Graders](https://docs.openreward.ai/environments/using-llm-graders.md): Use language models to assign rewards
- [Using Rubrics](https://docs.openreward.ai/environments/using-rubrics.md): Assigning rewards for harder-to-verify domains
- [Using Task-Specific Tools](https://docs.openreward.ai/environments/using-task-specific-tools.md): Define tools that are specific to individual tasks
- [Using the CLI](https://docs.openreward.ai/environments/using-the-cli.md): Manage environments, deployments, and runs from the command line
- [Using Toolsets](https://docs.openreward.ai/environments/using-toolsets.md): Create reusable tool collections and compose them into environments
- [Ways to Access Tasks](https://docs.openreward.ai/environments/ways-to-access-tasks.md): List, count, and fetch tasks from an environment
- [Where Environment Data Lives](https://docs.openreward.ai/environments/where-environment-data-lives.md): Understanding the difference between server-side data and sandbox-mounted data
- [Your First Environment](https://docs.openreward.ai/environments/your-first-environment.md): Make and deploy your first ORS environment
- [Harness Toolsets](https://docs.openreward.ai/harnesses/harness-toolsets.md): Expose agent-native tool surfaces in your environments using session-scoped toolsets
- [Harness Quickstart](https://docs.openreward.ai/harnesses/quickstart.md): Run agent harnesses against OpenReward environments with firehorse
- [What is OpenReward?](https://docs.openreward.ai/index.md)
- [Using Harbor Environments](https://docs.openreward.ai/integrations/using-harbor-environments.md): Convert Harbor tasks into OpenReward environments using harbor2or
- [Using Verifiers Environments](https://docs.openreward.ai/integrations/using-verifiers-environments.md): Convert Verifiers tasks into OpenReward environments using verifiers2or
- [Quickstart](https://docs.openreward.ai/quickstart.md): Sample from an environment in minutes
- [Recording Rollouts](https://docs.openreward.ai/rollouts/recording-rollouts.md): Record and upload agent rollouts to OpenReward for viewing and analysis
- [Using Daytona Sandboxes](https://docs.openreward.ai/sandboxes/daytona.md): Make and deploy an ORS environment with Daytona Sandboxes
- [Using E2B Sandboxes](https://docs.openreward.ai/sandboxes/e2b.md): Make and deploy an ORS environment with E2B Sandboxes
- [Using Modal Sandboxes](https://docs.openreward.ai/sandboxes/modal.md): Make and deploy an ORS environment with Modal Sandboxes
- [Using OpenReward Sandboxes](https://docs.openreward.ai/sandboxes/openreward.md): Use OpenReward native sandbox infrastructure with ORS environments
- [Why use Sandboxes?](https://docs.openreward.ai/sandboxes/why-sandboxes.md): Understand what sandboxes are and why environments need them
- [Cloud Storage](https://docs.openreward.ai/storage/buckets.md): Using cloud storage with environments and sandboxes
- [Debugging Training](https://docs.openreward.ai/training/debugging-training.md): Common errors encountered during training runs and how to handle them
- [Training with Miles](https://docs.openreward.ai/training/training-with-miles.md): Train language models with reinforcement learning using Miles and OpenReward environments
- [Training with SkyRL](https://docs.openreward.ai/training/training-with-skyrl.md): Train language models with reinforcement learning using SkyRL and OpenReward environments
- [Training with Slime](https://docs.openreward.ai/training/training-with-slime.md): Train language models with reinforcement learning using Slime and OpenReward environments
- [Training with Tinker](https://docs.openreward.ai/training/training-with-tinker.md): Train language models with reinforcement learning using Tinker and OpenReward environments

## OpenAPI Specs

- [openapi](https://docs.openreward.ai/api-reference/openapi.json)