Backstage Plugin · Open Source

AI Assets,
One Hub.

A centralized hub for AI Instructions, Agents, Skills, Workflows and Bundles — synced from Git, served via MCP, and available to every AI tool your team uses.

Works with your AI tools

Claude Code
GitHub Copilot
Google Gemini
Cursor
Dev AI Hub screenshot

Everything your team needs

From Git sync to MCP server, Dev AI Hub handles the full lifecycle of AI assets in your organization.

🔄

Git-Synced Assets

Sync from GitHub, GitLab, Bitbucket, Azure DevOps or any Git provider. Assets are always up to date. Schedule syncs or trigger them manually.

Embedded MCP Server

Built-in Model Context Protocol server using StreamableHTTP — no separate process needed. AI tools discover and install assets automatically.

🎯

Smart Install Paths

Per-tool path conventions auto-resolved. The right file ends up in the right place for each AI tool — Claude Code, Copilot, Gemini or Cursor.

📊

Usage Metrics

Track how many times each asset has been installed, via MCP or direct download. See which assets your team uses most.

🔍

Rich UI

Browse, search and filter assets by type, tool and tags. Markdown preview, raw YAML, metadata — all in one panel. "New" and "Updated" badges keep your team informed of fresh content.

🗄️

Backstage Native

Uses Backstage database service with automatic migrations. SQLite for development, PostgreSQL for production. Zero manual setup.

🗂️

MCP Server Catalog

Curate a list of third-party MCP servers in a mcp-catalog.yaml file in your assets repo. One-click install for VSCode and Cursor — no Backstage deploy needed.

🌐

Internationalization

All UI strings are translatable via Backstage's native i18n system. Included locales: English, Brazilian Portuguese and Spanish. Integrators can add extra languages in App.tsx.

Five types of AI assets

Each asset is a YAML envelope plus a Markdown file. Bundles group multiple assets into a single installable unit with a step-by-step wizard.

📄

Instructions

Persistent behavior rules applied to every conversation in a project. Equivalent to system prompts.

🤖

Agents

Specialized sub-agents with a defined role, model and behavior. Used in multi-agent orchestration.

🔧

Skills

Reusable capabilities following the open Agent Skills standard. Compatible across tools.

🔀

Workflows

Step-by-step processes that guide the AI through complex multi-stage tasks.

📦

Bundles NEW

Group instructions, agents, skills and workflows into one installable unit. Install step-by-step via wizard or download as a .zip.

Your AI tools, connected

Point any MCP-compatible tool at your Backstage instance. The ?tool= param filters assets per client.

search_assets Full-text search across name, description and tags
list_assets List assets filtered by type or tool
get_asset Fetch full metadata and pure Markdown content
install_asset Get content + install path; model creates the file
install_assets Batch-install multiple assets in one MCP call — prefer over N individual installs NEW
install_bundle Install all assets in a bundle at once — returns paths + curl commands for each file NEW
get_popular Returns the most-installed assets, optionally by type
list_providers Lists configured repositories with sync status and counts
suggest_assets Proactively suggests assets based on project context — opt-in via ?proactive=true
check_for_assets Prompt: triggers asset discovery before starting work — opt-in via ?proactive=true
.mcp.json
{
  "mcpServers": {
    "dev-ai-hub": {
      "type": "http",
      "url": "http://backstage:7007
        /api/dev-ai-hub/mcp
        ?tool=claude-code"
    }
  }
}
✦ Ask your AI tool
"List available agents in Dev AI Hub"
"Install the code-review skill"
"Search for security instructions"

Up and running in minutes

1

Install via npm

Run yarn workspace backend add @julianpedro/plugin-dev-ai-hub-backend and yarn workspace app add @julianpedro/plugin-dev-ai-hub in your Backstage monorepo.

2

Register the backend

Add backend.add(import('@julianpedro/plugin-dev-ai-hub-backend')) to your packages/backend/src/index.ts.

3

Register the frontend

Add devAiHubPlugin to the features array in your createApp(). Page, sidebar and API register automatically.

4

Configure providers

Point at your Git repo in app-config.yaml and set your GITHUB_TOKEN. Done.

app-config.yaml
devAiHub:
  providers:
    - id: main-ai-assets
      type: github
      target: https://github.com/
        your-org/ai-assets.git
      branch: main
      schedule:
        frequency:
          minutes: 30
        timeout:
          minutes: 5

integrations:
  github:
    - host: github.com
      token: ${GITHUB_TOKEN}
5
Asset types
4
AI tools supported
10
MCP tools
0
Extra processes needed

Ready to centralize your AI assets?

Open source, Apache 2.0 licensed. Drop it into your Backstage instance today.