We're shipping two new tools: InsForge Agent Skills and the InsForge CLI. Together they close the loop between AI-assisted development and hands-on infrastructure management.
Agent Skills
InsForge is built as an agent-oriented backend. That means AI coding agents -- Claude Code, Cursor, Windsurf, and others -- should be able to set up your backend, write integration code, and deploy your app without you copy-pasting from docs.
To make that work well, agents need structured context. That's what InsForge Skills provide.
Why Skills Instead of MCP?
InsForge already has an MCP server that works well. But Skills offer a lighter alternative with better performance for knowledge-heavy tasks.
MCP servers run as separate processes that communicate over a protocol. Every tool call is a round trip -- the agent sends a request, waits for a response, and then continues. For tasks that need external APIs or live data, that makes sense.
Skills work differently. They're structured markdown files that load directly into the agent's context. No subprocess, no network latency, no protocol overhead. The agent reads the skill and immediately has the knowledge it needs. For documentation, patterns, and workflows -- which is most of what backend integration requires -- this is significantly faster.
Skills also follow the Agent Skills Open Standard, developed by Anthropic and adopted by over 40 AI development tools including Claude Code, Cursor, GitHub Copilot, Google Gemini CLI, VS Code, JetBrains, and more. Write once, use everywhere -- the same SKILL.md files work across all of them.
What is SKILL.md?
A SKILL.md file is a structured markdown document with YAML frontmatter that tells an AI agent what it needs to know. Each skill lives in its own directory and can include supporting reference files:
skills/
├── insforge/
│ ├── SKILL.md # Main manifest and overview
│ ├── database/
│ │ ├── sdk-integration.md
│ │ └── backend-configuration.md
│ ├── auth/
│ │ ├── sdk-integration.md
│ │ └── backend-configuration.md
│ └── ...
└── insforge-cli/
├── SKILL.md # CLI command reference
└── references/
├── db-query.md
├── functions-deploy.md
└── ...
Agents discover skills automatically based on their descriptions. When you ask your agent to "add authentication to my app," it loads the relevant skill and has all the patterns, examples, and edge cases it needs -- without you pointing it to docs.
Two Skill Modules
insforge covers frontend SDK integration. When your agent needs to write a database query, add authentication, upload files, or call the AI gateway, this skill provides the patterns and examples it needs.
insforge-cli covers backend infrastructure. When your agent needs to create tables, deploy functions, manage secrets, or set up deployments, this skill guides it through the right CLI commands.
The separation is intentional. SDK code and infrastructure commands are different contexts, and agents make fewer mistakes when the boundary is clear.
Installation
# Skills registry (works with any compatible agent)
npx skills add insforge/insforge-skills
# Claude Code
/install-skills insforge/insforge-skills
Once installed, your agent automatically picks up InsForge context when you describe what you want to build. No manual doc lookups needed.
InsForge CLI
The CLI gives you direct control over your InsForge project from the terminal. Every infrastructure operation that the dashboard supports is available as a command.
npm install -g @insforge/cli
What You Can Do
- Database -- run SQL, inspect schemas, manage RLS policies, export and import data
- Functions -- deploy edge functions, view source, invoke endpoints
- Storage -- create buckets, upload and download files
- Deployments -- deploy frontend apps, track status
- Secrets -- manage environment variables for your functions
- Schedules -- create and manage cron jobs
Agent-Friendly by Design
Every command supports --json for structured output and -y to skip confirmation prompts. Exit codes are semantic -- agents can detect auth failures, missing projects, and permission errors programmatically. This makes the CLI work seamlessly in agent-driven workflows and CI/CD pipelines.
# Non-interactive, structured output
insforge db query "SELECT count(*) FROM users" --json
insforge functions deploy ./my-function -y --json
CI/CD Ready
The CLI supports environment variable authentication, so you can use it in GitHub Actions, GitLab CI, or any pipeline without interactive prompts.
Skills + CLI Together
The real power is in the combination. Your agent consults the skills to understand what needs to happen, then uses the CLI to execute it. You describe what you want to build. The agent figures out the infrastructure commands, writes the application code, and ships it -- all from a single conversation.
This is what agent-oriented backend development looks like. Less wiring. More building.
Get Started
Everything runs through npx -- no global installs required. Create a new project, install skills, or run any CLI command directly:
# Create a new InsForge project
npx @insforge/cli create
# Link an existing project
npx @insforge/cli link
# Install skills for your AI agent
npx skills add insforge/insforge-skills
# Run any CLI command without installing
npx @insforge/cli db query "SELECT count(*) FROM users"