Hubfly space for AI agents

Your agent can ship it.You stay in control.

Teach your coding agent to deploy with the CLI, create and connect Managed Databases, Realtime servers and Storage buckets, and use the Dashboard API for domains, HTTPS, resource configuration, and diagnosis.

Project workflow01 / 03
01
SourceCLI + Dashboard API
02
ControlDatabases, Realtime, Storage
03
ReleaseVerified deployments
Hubfly space projectNetwork · volumes · domains · team access
One project model

Install

Add the skill to your project.

The portable project path is the simplest default. Place the skill and its references in your agent's native skill directory. A standalone SKILL.md also links to the published references.

terminal

$ mkdir -p .agents/skills/hubfly-cli/references
$ curl -fsSL https://www.hubfly.space/skills/hubfly-cli/SKILL.md \
-o .agents/skills/hubfly-cli/SKILL.md
$ for ref in api networking diagnostics databases realtime storage analytics; do
curl -fsSL https://www.hubfly.space/skills/hubfly-cli/references/$ref.md \
-o .agents/skills/hubfly-cli/references/$ref.md
done

Agent products may use different discovery directories. The file itself follows the portable Agent Skills format and does not contain credentials or executable scripts.

Portable project

.agents/skills/hubfly-cli/SKILL.md

Shared convention for compatible coding agents

Codex

.codex/skills/hubfly-cli/SKILL.md

Project-scoped Codex installation

Claude Code

.claude/skills/hubfly-cli/SKILL.md

Project-scoped Claude Code installation

Cursor

.cursor/skills/hubfly-cli/SKILL.md

Project-scoped Cursor installation

Prerequisite

Connect the CLI and API.

The agent prefers the CLI for builds, deployments, logs, tunnels, managed databases, realtime servers, and storage. For additional operations it uses the Dashboard API with a securely supplied token and the live API contract.

hubfly-cli

$ curl -fsSL https://raw.githubusercontent.com/hubfly-space/hubfly-cli/main/install.sh | bash
$ hubfly version
$ hubfly login
$ hubfly whoami

Workflow

Inspection first. Verified result last.

The skill gives agents enough deployment knowledge to be useful while keeping infrastructure changes under your control.

01

Inspect

The agent reads the app, schema-v2 manifest, ports, volumes, resources, and environment-variable names.

02

Reconcile

It compares explicitly managed local fields with the Dashboard's bound cloud container.

03

Plan

It reviews the server plan, revision guard, and destructive operations before proposing a rollout.

04

Authorize

It follows your deployment scope and asks when a new cost, target, or destructive action needs a decision.

05

Follow

It tracks the asynchronous deployment session and reports the real terminal result or exact failed phase.

Try it

Ask in plain language.

The agent chooses the CLI or API for your goal, finishes public apps with HTTPS, binds managed services to your app without printing their secrets, and checks that the app reaches them before reporting success.

“Inspect this repository and deploy it to Hubfly space.”

“Compare my schema-v2 manifest with the bound cloud container and show the deploy plan.”

“Pull the current cloud configuration without exposing secret values.”

“Deploy this Compose stack and verify every service.”

“Diagnose this build ID from its deployment status and events.”

“Deploy my web app with HTTPS and keep Postgres private with tunnel access.”

“Create a managed Postgres database for this app and bind it.”

“Add live notifications to this app with Hubfly space Realtime.”

“Store user uploads in a Storage bucket with a key scoped to it.”

Safety boundary

No silent infrastructure changes.

The agent carries out the deployment you authorize, including required ingress. It shows the quote before creating a paid database, realtime server, or Storage project, and asks before expanding scope to purchases, paid upgrades, route transfers, public database exposure, credential rotation, or destructive data changes. Deleting volumes, instances, buckets, or objects requires explicit confirmation.

Secrets stay secret

The skill never asks an agent to commit tokens or print secret values. Database URLs, realtime secrets, and S3 keys go straight into your project through bindings.

--yes is not permission

The flag suppresses a CLI prompt; your request and authorized scope determine what the agent may do.

Sessions are authoritative

A local build or regional upload is not reported as success until Dashboard reaches a terminal deployment state.

Ready for your agent

Install the skill. Describe the outcome.

Hubfly space handles the deployment path while your agent inspects, configures, deploys, diagnoses, and verifies.