huggingface_hub 1.14.0 adds Space secrets management and pushes Hub automation further into the CLI
Hugging Face’s latest huggingface_hub release matters less for a single flashy feature than for how it keeps turning the Hub CLI into a real operations surface for Spaces, buckets, and agent-friendly workflows.
A lot of developer-tool updates are easy to ignore when they look like another round of command additions. huggingface_hub 1.14.0 feels more relevant because the new commands make the Hub more useful as an operational surface, not just a download client.
The headline addition is straightforward: you can now manage Hugging Face Space secrets and environment variables directly from the CLI. But the bigger story is that the Hub CLI keeps becoming a control layer for workflows that normally get split across dashboards, scripts, and manual cleanup.
What is actually confirmed
From the official GitHub release notes for v1.14.0:
hf spaces secretscan list, add, and delete write-only secretshf spaces variablescan manage readable environment variablesHfApi.get_space_secrets()exposes secret metadata without revealing valueshf buckets cpnow supports rsync-style trailing-slash behavior for folder copies- the old
hf skills upgradecommand has been renamed tohf skills update
PyPI independently confirms that huggingface-hub 1.14.0 was released on May 6, 2026 and is now the latest published package version.
The broader Hugging Face v1.0 blog post adds useful context for why this matters. The company describes the library as a mature foundation for Hub access, file transfer, inference-provider access, and agent tooling, with an explicit push toward more capable CLI and MCP-friendly workflows.
Why the Spaces secrets feature stands out
The new secret-management commands solve a practical problem: too many Spaces workflows still depend on clicking around a web UI when they should be scriptable.
That matters for:
- teams managing multiple Spaces
- creators iterating on demos with API-backed features
- automated deployment flows
- agentic workflows that need predictable environment handling
A CLI that can handle secret metadata, environment variables, bucket copies, and format-friendly output is much easier to plug into repeatable tooling than a browser-first admin flow.
The more interesting shift is about automation ergonomics
The release is not only about secrets.
The trailing-slash copy semantics for hf buckets cp, the continued expansion of CLI surface area, and the agent-friendly formatting direction all point in the same direction: Hugging Face is making the Hub easier to treat like programmable infrastructure.
That is useful even for people who never touch classic MLOps stacks. Smaller teams, solo builders, and product-minded developers increasingly want one place to:
- host demos and Spaces
- move files reliably
- manage runtime configuration
- connect model workflows to scripts and agents
The more of that lives in a stable CLI, the more the Hub becomes operationally sticky.
What to watch next
The next question is whether Hugging Face keeps deepening this operations layer around:
- safer secret rotation
- better auditability
- more Space lifecycle controls
- stronger agent-oriented command output
- smoother bridges between local scripts, MCP tools, and hosted Hub assets
If that continues, releases like 1.14.0 will look less like incremental polish and more like the quiet infrastructure work that makes agent-driven development feel normal.
Bottom line
huggingface_hub 1.14.0 is a practical release. It makes Space configuration more scriptable, improves file-copy behavior, and keeps the Hub CLI moving toward an infrastructure-grade tool for modern AI workflows.
That is not the loudest kind of AI news. It is often the more durable kind.