Documentation
Set up AI Chat and Pipeline Builder
Connect your own AI provider, choose models and start a chat or a pipeline draft. Your provider bills model usage directly.
Connect your provider
An owner or a member authorized to manage workspace AI opens Settings → AI, selects bring-your-own mode, adds an Anthropic or OpenAI provider key, and tests the connection. Choose a model from the provider’s model list. Keep the key in settings, never in a chat prompt or workflow file. Workspace AI permissions and configured token and cost limits still apply.
Use AI Chat
Enable the assistant in Settings → AI and choose its model. Open AI Chat and send a question, for example: What does my last failed job say? Available context depends on your workspace role and connected repositories. Runzivo adds no AI usage charge. Your provider bills model usage directly.
Build a pipeline draft
Pipeline Builder is available to every workspace. In Settings → AI, enable the Pipeline builder agent and choose its model. Open Pipelines under CI & Runners, select an authorized repository and describe the workflow you want. Without a configured key, the page directs you to AI settings.
The builder replies with a draft. Repository writes stay off in production. There is no commit or pull request button. Review the proposed workflow and apply it through your own repository review process. AI Fix has separate workspace availability.
Run a sandbox coding task
Choose New coding task in AI Chat and select a sandbox, repository and model. Follow the conversation, commands, file changes and tests, then send another message to continue. Sandbox compute uses the normal workspace allowance or credit. Disk charges are currently off.