beehiivYou have Claude Code running in three terminals and Codex on a server. Twenty minutes ago one of them stopped to ask whether it could drop a column. It’s still asking.
You find out when you happen to click on that window. Meanwhile the session you left on the landing page has used most of your five-hour window, and you’ll find that out when it stops too.
Running agents isn’t the hard part anymore. Keeping track of them is. Context Board gives them one screen.
What the Agents board shows
Every new account gets an Agents board beside its first one. Link your computer and it fills in.
- Agent activity: every Claude Code and Codex session on your computers, with its title and project. Waiting for you comes first, then working now, then done in the last hour.
- Needs you: everything waiting on you. A session that asked a question, wants your OK to use a tool, or has a plan for you to approve. Jobs that failed land here too.
- AI usage: what’s left of each plan’s current five-hour window and its week, as rings that turn amber under half and red under a fifth, each with when it resets. More on the Claude Code usage and Codex usage pages.
- Git activity: today’s commits, the branch and what isn’t committed yet, in the project folders you choose.
- Morning brief: written by your own agent from your board, at the time you pick.
The sessions come from wherever you run them. A terminal, VS Code, or an app that runs Claude Code or Codex for you.
How sessions get on the board
A small program called the runner links your computer to your board. In Connect, open Your AI, then Your computers, and copy the one line for Windows or for Mac and Linux. Run it once. It brings Node.js itself if the computer has none, and it starts with the computer from then on.
Do the same on your laptop, your desktop and any server where agents run. Sessions, usage and git come from all of them.
From your sessions, only names, states and percentages leave each computer. A session’s title, whether it’s working, how much of a limit is left. Never what you and your agent wrote.
Jobs your agents run for you
The runner also lets the board use your own Claude Code or Codex, on the plan you already pay for. No API key.
So the board can put your agents to work. Turn on the morning brief and it’s written at the time you pick, with the one thing that matters today, what changed, what needs you and the three things to do first. Or give a Number, Chart, List or Table widget a job in your own words, like “how many TODOs are open across my project folders”, and your agent runs it every day, every week or when you press Run now. Only you give a widget its job or set a schedule. An agent can write a brief, never set one to run.
Agents that work on the board itself
The board is also an MCP server. Connect Claude, ChatGPT, Cursor, Claude Code, Codex or VS Code by its address, click Allow, and those agents can read the board, send it data, build widgets and post what they finished. Each update shows with the agent’s name on it, and anything it changes can be undone.
Put revenue or GitHub beside your agents and you see the work and what it’s for, on one screen.