Your engineers have assistants now. You cannot see any of it.
The work happens in someone's terminal, the reasoning lives in a chat window only they can read, and the first you hear about it is standup. It is also not one tool: one engineer is in Claude Code, one is in Cursor, one is driving Codex. Maestro Mojo puts all of it on a board your whole team already has open — and marks which rows an assistant wrote, and which assistant.
Written as of August 2026
Agents file, claim, comment and move items on the same board your team opens. Changes patch in live — nobody refreshes, and nobody reports.
A row an agent writes appears on an already-open board immediately. The person watching does not have to know an agent was involved to see the work.
The board belongs to the workspace, not to a repo. Work that crosses projects — which agent work usually does — stays on one surface.
An assistant acting through Maestro authenticates as the person using it, so the author stays human — and the row carries a small badge naming the tool behind it.
Your team will not standardise on one assistant, and it does not have to. Claude Code, Cursor, Codex, Windsurf, Cline, Zed — they all speak MCP, so they all write to the same board, and each row says which one wrote it.
Anything you type in Workspaces yourself carries no badge at all. That is what gives the badge meaning: its absence is a statement too.
Seen on one board via Claude Code · via Claude Desktop · via Cursor · via Codex · via Windsurf · via Cline · via Zed · via any MCP client
Worth stating precisely, because it is the part people ask about: the badge is set by the server and cannot be forged — an API client that tries to attach one has it stripped — and anything you type in the portal yourself never carries one. That is what makes its presence mean something.
The name in the badge is a different matter: the client reports its own name, so “via Cursor” is a claim about which tool, not proof of it. The guarantee is one-directional and we would rather say so plainly: a badge is never faked onto a human's own row, and nobody can quietly launder assistant work as their own.
Every number in a digest is counted out of the database over an explicit window. The model only writes the sentences around them.
The most valuable thing an assistant produces is usually not the code. It is the reasoning about why the work is shaped that way — and that is the part that disappears.
The item carries the spec, the plan, the commits and every comment along the way. Close the terminal, change laptops, hand it to someone else — it is all still there.
A chat transcript belongs to whoever opened it. A board item belongs to the workspace, so the next person — or the next agent — starts from what is already known.
Each item leads with plain language: the problem, why it matters, what done looks like. The technical detail sits below it for whoever is building.
The point is not surveillance. It is that work done by assistants should leave the same trail as work done by people — so a manager can read it, a teammate can pick it up, and nobody has to reconstruct a decision from a chat log they never had access to.
One person on your team connects it in a couple of minutes; everyone else just opens Workspaces and looks. Nobody has to change the assistant they already like.
Works with Claude Code · Claude Desktop · Cursor · Codex · Windsurf · Cline · Zed · any MCP client
Free when you bring your own AI — credits only apply when you use Maestro's AI
Maestro Mojo