From request to result
- A teammate starts a request from the web dashboard, a communications provider, source control, Linear, an automation, or the API.
- Roomote creates or resumes a session, chooses the model and any authorized integrations, and selects an environment when the request needs a workspace.
- A task runs in an isolated sandbox with the repositories, services, and variables defined by that environment.
- The task records its conversation, terminal output, logs, artifacts, code diff, and live preview as it works. The session keeps the explanation and reports from delegated tasks together.
- Roomote returns the result to the originating surface and, for code changes, can deliver a branch, pull request, or reviewable diff.
What operators configure
- Inference providers supply the models used for coding, orchestration, vision, review, and helper work.
- Environments define repositories, services, variables, and the sandbox setup a task can use.
- Communications providers and source-control providers connect Roomote to the places where people start work and review results.
- Mediated built-in integrations provide additional authorized context and tools while keeping their credentials outside task sandboxes.
- MCP servers depend on their transport and configuration: remote custom
servers use Roomote’s proxy, while environment-defined servers and
deployment-scoped custom stdio servers can deliver configured HTTP
headersor processenvvalues to the task sandbox that runs them. - Automations start repeatable sessions from schedules, provider events, manual actions, or webhooks.
Where to continue
- Start with How Roomote works for surfaces, sessions, and environments.
- Configure an operator deployment with Self-hosting and Environments.
- Choose execution capacity from Sandboxes and inference from Models.
- Review evidence in Sessions and Tasks.
- Use Local development only when you intend to modify Roomote itself.