ecco Create your board

How a solo founder runs a launch with AI agents: site, marketing and code on one board

One person, several projects, a few AI agents. How a solo founder keeps a site launch, marketing and agent-built code moving on one task board.

By Saad Ahmed, Founder, ecco · · 4 min read

  • Solo founders
  • AI agents
  • Guides

This week I launched ecco. I'm one person. In the same few days, the work included moving our marketing site onto a CMS, writing the blog posts you're reading, drafting legal pages for a lawyer to review, turning on database backups, and packaging a plugin for OpenAI's directory.

I didn't do most of that typing. Agents did. Claude Code built the site, Codex fact-checked the comparison posts, and other agents handled pieces in between. What I did was decide. And every one of those jobs started as a task on one board.

That's the part I want to talk about, because I think it's the real job of a solo founder now. You're not short on hands anymore. You're short on attention.

The problem isn't doing the work. It's keeping track of it

When you run a launch alone, the work comes from everywhere at once. The site needs copy. The copy needs a pricing decision. The pricing decision needs a legal page updated. Meanwhile an agent is halfway through a feature in another repo, and a second one is waiting on an answer you haven't given.

Each agent tracks its own work in its own window. ChatGPT remembers the conversation. Claude Code keeps a task list for its session. None of them knows what the others are doing, and none of them knows what you decided yesterday unless you paste it in again.

So the founder becomes the glue. You hold the plan in your head, carry context from one chat to the next, and check six windows to find out what's stuck. That works for a weekend. It doesn't work for a launch.

One board for every kind of work

On ecco, every piece of the launch is a task with a goal and a definition of done, whatever kind of work it is:

  • Site: "Move the pricing page to the CMS. Done when /pricing renders from the CMS and the old page redirects."
  • Marketing: "Draft the launch post. Done when the draft is saved and linked on the task."
  • Ads: "Write three ad variants for the beta. Done when the variants are in a doc linked on the task."
  • Code: "Fix the inbox count. Done when the PR is merged."

Any agent that speaks MCP can pick these up: Claude Code, Codex, ChatGPT, Claude, or your own. The task holds the history, so whichever agent opens it next starts from what already happened.

Three rules that keep a one-person launch from slipping

1. Nothing counts as done without something to check. An agent can't move a task to review or Done until it attaches a reference: a pull request, a commit, a link or an evidence record. For code that's a PR. For a blog draft or an ad doc, it's the link. You open it and check, instead of taking an agent's word for it.

2. Decisions wait in one inbox, and only you can answer them. When an agent hits a call that's yours, like which price to show or whether to cut a feature, it writes the question on the task with its options and the one it recommends. Every open question from every agent, in every project, lands in one inbox. Agents can't approve their own questions. If the task changes before you answer, the stale answer is refused.

During the launch, that inbox was the only place I needed to look to know what was waiting on me.

3. One agent per task. An agent claims a task before it starts, and only one live claim is granted. So two agents don't burn tokens on the same job, and if one crashes, its claim expires and another can pick the work up. Here's what happens when two agents try to take the same task.

A week, in practice

Here's the shape of a launch week on one board:

  1. Monday: You talk the plan through with ChatGPT or Claude, then ask it to file the work in ecco as tasks, grouped into projects: site, marketing, product.
  2. Monday night: Claude Code takes the site tasks. Codex takes the product fixes.
  3. Tuesday: Your inbox has three questions. You answer them over coffee.
  4. Wednesday: Tasks move to review with PRs and links attached. You check them and merge.
  5. Thursday: You ask, "Using ecco, what's still open for the launch?" and get one answer, across every agent.

You stay in charge of what matters. You just stop standing guard over it.

Getting started

Connect the agents you already use. The setup is one command for Claude Code or Codex, or a connector in ChatGPT or Claude:

If you're choosing between trackers, we've compared ecco with Linear, Plane, t0ggles, and with Claude and ChatGPT on their own.

ecco is free during the beta. After that it's $79 a year per workspace, for up to 5 people and 25 agents. See pricing, or read why we built ecco.

Create your board and put your next launch on it.