Your agent writes the code. Trood takes care of everything around it: technology choice, security, quality, putting the project online, the systems you run, and the people who can help. As much help as you want, area by area.
Works with Claude Code, Claude Desktop and any MCP client. Free. Your keys and your code never leave your computer.
So you are backed with top tech expertise automatically. Trood's cookbook is the stack and the practices we run in production ourselves: how a product is structured, how it stays secure, how it is tested, how it goes live. Tell your agent what you are building, and it adapts the cookbook to your industry and product.
Your agent builds on that instead of guessing. You get a product that holds up, without having to know how it all works underneath.
From the first sentence about your idea to a tested product people use: where the code is kept, where the app runs, and a test version for your team when you want one. Your agent does the work; you approve a couple of sign-ins and see every price first.
Guide, hint or quiet, separately for deploys, secrets, domains and the rest. Say "stop explaining deploys" and it stops. A no is remembered.
Tell Trood what you need in your own words, and get a mentor call when you are stuck. Soon, a gap the checks find will become a project with the right person on it.
For studios and agencies: research a company's website, record facts with their sources, and draft a precise offer, all on your own model. Trood pays for no AI and sees none of your research.
A map of what runs where, across repositories and vendors: owners, environments, where each key lives and when it expires. Never the keys themselves.
Weekly, for what you run only: security and quality checks, keys still valid, nothing about to expire, new versions and security advisories. Quiet unless something needs you.
Search Trood's Technology Tree today. Soon, every choice and its reason will be saved on the project, so the next person and the next agent know why.
Architecture notes, decisions, runbooks and business documents will live on platform.trood.com, read by your agent in any repository. Export them any time, or mirror them to your own git.
The same request, at the three levels. You set the level per area, or just tell your agent.
One command in Claude Code, or a few lines in any MCP client.
One question: do things with you, or just keep an eye on things. It sets the levels and offers a short block for your CLAUDE.md.
Talk to your agent as usual. It reaches for Trood when the question is about the environment, a person or a tutorial, and writes the code itself.
The server, the /trood:start and /trood:launch commands, and a skill that routes environment questions to Trood.
/plugin marketplace add TroodInc/trood-mcp
/plugin install trood@troodJust the MCP server.
claude mcp add trood -- npx -y @troodinc/mcpIn claude_desktop_config.json.
{
"mcpServers": {
"trood": { "command": "npx", "args": ["-y", "@troodinc/mcp"] }
}
}Runs over stdio.
npx -y @troodinc/mcpOptional: sign in with your Trood account to connect your workspace on platform.trood.com.
Your credentials, your research and your settings stay there, readable by you only.
Your own model does the thinking. Trood never receives your code.
Usage is recorded only after you say yes: tool names and whether they worked, never what you typed. See everything it sent, any time. Do Not Track always wins.
It never sends messages, approves, pays or changes anyone's access. A person does those, on platform.trood.com.
Apache License 2.0. Read every line.
No. Claude reasons and writes the code. Trood holds what no model can know from your repository: what runs where across vendors, live checks with your own credentials, a weekly watch of your stack, and people we vouch for.
The MCP is free and open source. Trood earns when you ask for human work, which becomes a project with fixed stages. Hosting and other services charge you directly, on your own accounts, and you see every price before you pay.
No. Launching, help levels, the Technology Tree and telling Trood what you need all work without one. An account connects your workspace on platform.trood.com.
Yes. Run /trood:start and choose to do things together. The agent explains each step in plain words, does the work, and waits for you only where a person is needed. Short tutorial videos are there when you want them, and a Trood mentor is a call away.
Nothing, unless you say yes. Then: which Trood tool was used, its area and help level, whether it worked, how long it took, your agent and its version, your OS family and the MCP version. Never your arguments, files, paths, repository or company names, code or documents.
A single platform for business owners, managers, technology and service providers to build the New World companies together.
Start with your agent. Bring the people in when you need them.