Storage, proxies & LLMs

Everything on this page can also be managed in the dashboard; the CLI just puts it in your terminal. Storage credentials, proxies and LLM credentials belong to your organisation. You create one once, then assign it to whichever projects should use it. Every group works the same way: list, show, create, update, delete, and an assign/unassign pair under projects.

Targets accept a name or an id (store_…, pool_…, llm_…). Secrets print masked everywhere; pass --reveal to unmask. --yes skips every prompt for scripts and agents.

Storage credentials

Send renders to your own bucket instead of Urlbox's hosting:

urlbox storage create prod --provider aws_s3 --bucket my-renders --region us-east-1 --key AKIA… --secret …
urlbox storage list
urlbox storage show prod --reveal

By default Urlbox hosts your renders for you; a storage credential saves them straight into a bucket you own instead. Saving to S3 explains the feature, with step-by-step setup guides for S3, Cloudflare R2, Google Cloud Storage, Backblaze B2, DigitalOcean Spaces and Azure Blob Storage.

The flags mirror what those guides walk you through:

• --provider picks the service: aws_s3, google_cloud_storage, cloudflare_r2, backblaze_b2, digitalocean_spaces, wasabi, custom, or azure.
• --bucket, --region, --key and --secret name the bucket and the access credentials Urlbox writes with.
• Azure takes --account-name, --container-name and --sas-token instead of a key and secret.
• --endpoint sets a custom endpoint for S3-compatible services, --private-bucket and --object-lock describe how your bucket is configured, and --cdn-host is your CDN's hostname if you serve the bucket through one.
• --json accepts the whole configuration as a single object if you'd rather not use flags.

Proxies

Your proxies route renders through servers you provide, which changes where in the world a render appears to come from and helps with sites that block datacenter traffic. The proxies guide explains when and why you'd use one, and avoiding being blocked covers the wider topic. Pass each proxy URL (credentials included):

urlbox proxies create eu-proxies --url 'http://user:pass@proxy1.example:8080' --url 'http://user:pass@proxy2.example:8080'
urlbox proxies show eu-proxies

Lists show a URL count rather than the URLs themselves, and show masks the password portion of each URL. update replaces the whole URL list with what you pass.

LLM credentials

An LLM credential connects your own AI provider account to Urlbox, powering the use_llm and llm_prompt render options that analyse a page as part of a render. You bring the model; Urlbox brings the page:

urlbox llm create extraction --provider openai --api-key sk-…
urlbox llm test extraction     # checks the credential against the provider
urlbox llm models extraction   # lists the models it can use

--provider can be openai, anthropic, azure, amazon-bedrock, google-vertex and more; cloud providers take their own fields via --json. The provider is fixed once created.

Assigning to projects

A credential does nothing until a project uses it:

urlbox projects storage assign my-project prod
urlbox projects proxy assign my-project eu-proxies
urlbox projects llm unassign my-project

Or do both steps at once when creating: urlbox storage create prod … --assign-to my-project. On a terminal, create also offers to assign interactively.