by devforth
Admin panels for humans and AI agents. Let agents search, update, and act on your data through built-in chat or MCP, with the same permissions as your team. Open-source and self-hosted.
# Add to your Claude Code skills
git clone https://github.com/devforth/adminforthGuides for using ai agents skills like adminforth.
Last scanned: 5/27/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-05-27T08:03:55.878Z",
"semgrepRan": false,
"npmAuditRan": true,
"pipAuditRan": true
}See how adminforth compares with popular alternatives.
adminforth is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by devforth. Admin panels for humans and AI agents. Let agents search, update, and act on your data through built-in chat or MCP, with the same permissions as your team. Open-source and self-hosted. It has 391 GitHub stars.
Yes. adminforth passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/devforth/adminforth" and add it to your Claude Code skills directory (see the Installation section above).
adminforth is primarily written in TypeScript. It is open-source under devforth on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh adminforth against similar tools.
No comments yet. Be the first to share your thoughts!
From $49 · 30 days · terms & refunds
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM lists this project for discovery and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
Automated security scan — not a certification.
The back office your coding agent builds and your agents run. Your data, your server, your rules. A change history that names the agent.
Website · Start with your agent · Documentation · For developers
You already work with AI agents. AdminForth is the admin they run your business through: customers, orders, jobs, stock and invoices in your own database, built by your coding agent, used by you and your team through Claude Code, Codex or any agent that speaks MCP. The panel is one way in; your agent is another. Screens are there when you want to look.
8 of the 10 biggest months for software vulnerabilities ever recorded happened in 2026. One in five of all vulnerabilities ever registered were published this year, and the year isn’t over (monthly data). Exploitation in the wild nearly doubled. An app your agent generated last month, with hundreds of packages nobody updates, sits right in that path. AdminForth, one framework maintained by our team, doesn’t.
my-business, and open the agent in it.MY BUSINESS BRIEF to your own process first.The agent builds the first part with sample data, runs the app, connects itself to it through MCP and shows you one operation end to end. From then on, you talk to your business.
Help me build a small business application with AdminForth. Work with me in
plain language. I don't have a developer: tell me exactly what to do myself,
step by step.
MY BUSINESS BRIEF
BUSINESS_PROCESS: Customer requests for my maintenance business.
CURRENT_PROBLEM: We send quotes and forget to follow up.
TEAM_ACCESS: Me and two managers. Each manager handles their own requests.
CURRENT_DATA: A spreadsheet. Start this example with synthetic data.
WHAT I WANT
A web app on AdminForth that my agent and my team's agents work with through
MCP, with sign-in, permissions and an audit log. I will mostly talk to it
through my agent; screens are for checking. Keep the first version small and
ready for more parts later (orders, jobs, stock, invoices).
START HERE
Read https://devforth.io/admin/api/af-start/business/ first, then
https://adminforth.dev/docs/tutorial/gettingStarted/ and
https://adminforth.dev/llms.txt. Use AdminForth as the framework: reuse
its screens, forms, filters and access rules, don't build a separate stack.
Use pnpm and Node.js 24. For a new local example use SQLite, the documented
create-app flow and Prisma migrations. Read the generated AGENTS.md, CLAUDE.md
and skills before changing anything. If something can't run here, tell me what
I must do instead of pretending it works.
Show me a short plain-language outline (records, roles, the first operation)
and ask only about decisions that change the data or the access rules.
BUILD
1. Records and relationships for my process, named so my team understands
them, with a small set of fictional sample records.
2. Access rules on the server: I manage everything; every other person works
only with what their role allows, including direct record requests, lists
and relationships. A hidden button is not an access rule.
3. The MCP plugin (https://adminforth.dev/docs/tutorial/Plugins/mcp/) so I can
connect my agent, and the Audit Log plugin
(https://adminforth.dev/docs/tutorial/Plugins/audit-log/) with its
executed-by column, so every change names the person and the agent.
4. No production data, emails or payments in this first version.
SHOW ME
Run the app and give me the local address and sign-in. Connect my agent with
the command from the Agents settings page, then do one operation through the
agent: find a record that needs attention, add its next action, and show me the
audit log entry. Show one action my agent is allowed and one a teammate's agent
would be refused. Say which checks you actually performed.
Write five prompts for my daily work with this app, in the words of my process.
Explain where the code and data are, how to reopen the app tomorrow, how to
update AdminForth and its plugins later, and what remains before my team uses
it: hosting, real data, backups, access review.
Prompts for the first day, with the example above:
What needs my attention today?Which follow-ups are overdue, and whose?Add a request from Harbor Café: office maintenance quote. Owner Maya, next action: send the quote by Friday.Which invoices are unpaid for more than 30 days?[!NOTE] The framework is free and MIT-licensed. Coding-agent access, model usage, hosting and paid services you connect have their own costs. Check each change in the app before your team relies on it.
⭐ Star AdminForth to find it again when your agents need an admin to work in.
AdminForth is a TypeScript and Vue 3 framework for back-office and business applications. Declare resources over your database and get list, show, create and edit screens with filters, validation and access control, plus the same resources as MCP tools for agents. Everything else is code you own: