Run parallel coding agents in cloud sandboxes and review their changes with Hoplite.
Hoplite runs coding agents in cloud environments connected to a software project. The workflow begins with a GitHub repository and can bring across local configuration, tools and agent context. Each session receives an isolated sandbox where the agent can work on code and run the project.
The official site describes importing sessions, MCP servers, memories and command-line tools. That can reduce repeated environment setup, but credentials and configuration should still be reviewed before being copied into a remote environment. Only the resources needed for the task should be available to the agent.
Parallel sessions allow different tasks to proceed separately. Hoplite advertises recordings and live previews that help reviewers inspect user-facing changes. A video is useful evidence of behaviour, but it does not replace tests for logic, permissions or failure cases that the recording does not show.
GitHub, Linear, Slack and MCP are named integration routes. Tasks can be handed between engineers and stakeholders through shared threads. Approval controls can gate edits, commands and pull-request actions according to the chosen configuration.
The free plan has project and concurrency limits and requires a card. Pro is priced per seat with included usage credits, while enterprise deployments use custom terms. Model use and sandbox allowances should be considered alongside the seat price.
Start with a small, noncritical change and inspect the resulting diff, tests and preview. Verify that secrets are scoped correctly and that approval behaviour matches your expectations. Keep a person responsible for merging and deployment until the team has established reliable review and recovery procedures.
