camelAI

What it is

camelAI is an AI coding agent that builds full-stack apps, admin panels, dashboards and scheduled automations from a conversation, then publishes them to a live URL. It is aimed at teams and founders who have data in a database and want an internal tool on top of it without writing the boilerplate. The company's site calls itself "a Y Combinator company making AI tools for people who build software," and says it started in 2022 "as three engineers from UT Dallas" and went through Y Combinator in Winter 2024. The pages do not name a headquarters. The coding product is called camelCode, it is MIT licensed, and you can run it yourself or use the hosted version.

What you build and how

The workflow is stated in three steps on the MySQL page: explore, build, ship. When you connect a database, camelAI "reads your INFORMATION_SCHEMA," meaning tables, columns, types, indexes, foreign keys and constraints, so it understands the database before it writes a query. Then "describe what you need in plain English. The agent writes the code, you review it." The sample prompt on that page is "Build an admin panel for the users table with search, plan filters, and inline editing." The last step is "One click to a live URL. SSL, CDN, and auth handled automatically," and the example output is an address on the camelai.app domain.

That makes internal tools the natural use. The MySQL page suggests four prompts: an admin panel for a users table with search by email, filters by plan and status, inline editing and CSV export; a customer data explorer that joins users, orders and products and shows purchase history timelines and total spend; a legacy database modernization dashboard that flags tables without indexes and unused columns; and a cron job that checks new orders every hour and posts a running revenue total to a Slack channel. The page positions camelAI against phpMyAdmin, MySQL Workbench, custom CRUD scripts and Metabase, and lists Retool at "$10/user/month" as the tool to replace. It says camelAI connects to "every flavor of MySQL": MySQL 5.7 and 8.x, MariaDB, Amazon Aurora, PlanetScale, Vitess, TiDB and Percona, plus any database reachable by a standard mysql:// connection string. The home page's connections menu also names PostgreSQL, MongoDB, BigQuery, Supabase and ClickHouse, with a "View All" link for more.

The coding agent itself is described on the home page as "a coding agent with a permanent computer." camelCode "works on a computer that keeps its files and databases between sessions," builds full-stack apps and scheduled automations, and publishes them. The home page's demo project is a revenue dashboard with an app folder, a Stripe file in the API folder, a database file and a daily report job scheduled for 9:00, published at an address on camelai.app. The open-source page shows the same idea as a chat: asked for a customer portal where clients upload files, track progress and message the team, the agent says it will create the portal, data model, authentication, file storage and project messaging, then deploy it to a live URL.

The use-case menu names AI App Builder, Data Analytics, Web Development, AI Agents, Automation, Email, Slack and Internal Tools. A second product, camelRun, is "cloud hosting for AI agents" at "$0.01 an agent-hour." You define an agent's model, instructions, tools and channels such as Slack, email and GitHub, and camelRun runs it, keeps its history and "scales to thousands of agents at once." Each agent sleeps until there is work, and the hourly rate applies only while it runs. The launch notes list camelRun SDKs for TypeScript and Python and drop-in React chat components, both in September 2026.

Open source is the part the company stresses most. The page states "Open source under MIT" and says "the interface, agent orchestration, persistent workspaces, build system, integrations, and publishing flow are available in the public repository." Self-hosting means deploying the stack "to a machine or private network," where you control "the environment, authentication, AI provider, backups, and upgrade cadence." The guided setup is three commands, bun run selfhost:init, bun run selfhost:doctor and bun run selfhost:up, which start a Docker Compose stack. Supported model providers are OpenAI, Anthropic, OpenRouter, Amazon Bedrock or a custom compatible endpoint. The page describes the architecture as a React Router interface on Cloudflare Workers and Durable Objects, with R2 and sandbox containers for build and analysis. Hosted camelAI is the other route: the company manages infrastructure, updates, workspaces and publishing, in the browser.

What the pages say about plans, license and claims

Prices are thin. The home page says camelCode has a free plan and "the free plan needs no credit card." The open-source page repeats "Free plan available" for the hosted product and uses the line "Try truly free vibe coding." The pages do not give paid plan names, prices, credit allowances or project limits for camelCode; a pricing page exists in the navigation. The only price figures are camelRun's $0.01 an agent-hour and the $10 per user per month the MySQL page attributes to Retool.

The license is the clearest claim: MIT, with the source in a public GitHub repository, and the page tags the product MIT licensed, self-hostable and provider flexible. The FAQ headings ask whether you can modify camelAI or use it commercially, but the page does not show the answers, so check the repository license for the exact terms. The site links a Trust Center, terms and privacy policy, but no compliance certification is stated.

The pitch behind the open-source release is control. In the company's words, "AI coding tools see the systems your business runs on. We believe you should be able to understand that software, control where it runs, and change how it works." Self-hosting is aimed at cases where "your code, credentials, and business context need to stay in your environment."

Who it suits and who it does not

camelAI suits an operations or data team with a MySQL, PostgreSQL or similar database and a backlog of admin screens, reporting dashboards and Slack digests nobody has time to build. It suits a founder who wants a quick internal tool and is comfortable reviewing what an agent writes. It suits a company that wants to read the code of the AI tool it points at its data, since the whole product is on GitHub, and an IT team that wants to choose its own model provider and keep the stack on a private network.

The limits are plain. This is a coding agent, not a drag-and-drop editor: "the agent writes the code, you review it" means the output is real code and someone should read it. Self-hosting is "designed for a single machine or private network," and "production DNS, TLS, networking, and operations remain under your control," so you take on the operations work. The pages publish no paid plan prices for camelCode, so a team that needs a budget figure has to look at the pricing page or ask. And camelRun is new, with its SDKs and chat components announced in September 2026.