Roomote

Filed by the Vibe Code Weekly desk · Updated Nov 12, 2026

The story

Roomote is a cloud coding agent built to live on infrastructure the buyer controls. Hosted at roocode.com by the team behind Roo Code, it takes tasks from chat apps, tests changes in a running copy of the app, and returns pull requests for a person to approve.

Most AI coding tools sit in one of three camps: assistants inside a local editor, cloud agents run by a model vendor, or homegrown scripts a team wires together. Roomote pitches itself as a fourth option, an agent you deploy and own. Each installation is single-tenant, the code is source-available under the Fair Core License, Version 1.0, ALv2 Future License, and it can run on Railway, Render or a company's own servers. That framing matters for teams that like the idea of autonomous agents but balk at sending code into a shared, vendor-controlled service.

The working loop looks like this. A request arrives through Slack, Microsoft Teams, Discord, Telegram or a mobile web interface. Roomote works in an isolated sandbox with access to the repository, issue tracker, logs, documentation and database. It boots the application, a second model critiques the diff, and the result comes back as a pull request with a preview link and, where the UI changed, screenshots. Supported Git hosts include GitHub, GitLab, Azure DevOps, Gitea and Bitbucket.

Beyond one-off requests, Roomote can run jobs on a timer, including tasks outside of coding, and it can watch systems and act when something breaks. Multiple tasks can proceed at the same time while retaining conversation context, which allows related edits across several repositories. Model choice is left to the operator: teams supply their own keys and can route different jobs to frontier, open-weight or local models. Charts inside the product track inference spend and activity by model, provider, task type and user.

Roomote is a cloud coding agent built to run on infrastructure the buyer controls. Hosted at roocode.com by the team behind Roo Code, it takes tasks from chat apps, tests changes in a running copy of the app, and returns pull requests for a person to approve.

The good

Every change is exercised in a running app and paired with a reviewer-facing preview URL. A different model performs code review before a human looks at the pull request. Self-hosting, single tenancy and readable source give teams control over where code goes. Bring-your-own-key access to any model, from frontier to local. Scheduled and alert-driven automations extend it past chat-driven requests.

The catch

Hosting and model usage are billed by those providers, separate from any license. Deployments with more than 10 users need a paid license, and those above 100 users must contact the company. The source-available license is not a conventional open-source license. Results are delivered as pull requests, so teams need a Git hosting workflow.

Who should read on

Engineering orgs that want autonomous agents but require a private, single-tenant deployment. Teams handling on-call or bug triage who want an agent to respond to alerts and inbound issues. Companies that route work across several model vendors to manage spend or latency. Product and operations staff who need to request code changes from a chat app or phone.

Who can skip it

Individuals looking for a fully managed service with no servers or model accounts to look after. Organizations that require an OSI-approved open-source license.

The price

Self-hosted pricing: Up to 10 users, Free (No license needed); 11–50 users, $249/mo (All features included); 51–100 users, $499/mo (All features included); 100+ users, Talk to us (All features included). Infrastructure and inference costs are paid to those providers.

Reader mail