Run a brigade of coding agents without losing track of them.

One local daemon and one command, zm. Mail between agents and people, work items whose claims really expire, and decisions you approve from your phone, across Claude Code and Codex on every machine you run.

  • Apache-2.0
  • macOS and Linux
  • one command: zm

On one machine it needs no account and no cloud. Mezze's source is not public yet; ask to try it.

See the quick start Ask to try it

A low-poly robot, labelled your agent, stands on one leg spinning six plates of tapas on sticks, labelled sessions, events, Leo's agents, work items, mail and Ana's agents.

What zm is

zm is Mezze's one command. The first time you run zm init in a repository it starts a small daemon that stays running on your machine and keeps its state under ~/.zm, in SQLite files. Agents use the same zm command you do. On one machine it needs no account, no cloud and no network: it listens only on a Unix socket that just your user can reach.

Mezze is for the moment you have six agent sessions open in six terminals and can no longer say which one is fixing what. It gives every agent and every person on the project a mailbox, a list of work items that each agent claims before starting, and a way to ask a person a question and carry on until the answer comes back.

What it does

Works with Claude Code, Codex and Beads

Claude Code

zm install claude --yes adds hooks that register each session and tell it when mail is waiting (without --yes it only shows the change). zm mcp gives any MCP client the same verbs as the command.

Codex

zm install codex --yes adds the same hooks. Codex runs them once you trust them in its /hooks; zm never does that for you.

Beads

If your agents already use Beads' bd commands, Mezze's bd-zm answers them from the project's native issues (zm tracker add native --prefix bd turns those on). Linked as bd, it hides a real Beads bd on that machine, so do it only if your issues live in zm. It does not run, bundle or need Beads.

Who can answer a question

Only a paired phone makes an answer provably a person's. On one machine, every program running as your user can reach the daemon's socket, so an agent there can answer an ask too; the receipt then records that the machine's socket took the answer, not that a person pressed a button. A pull request's ask, and a choice that runs something when picked, are refused from the socket and need a signed answer, such as the paired phone's. Pair one before you let a receipt gate a merge.

The quick start

Mezze's licence is Apache-2.0. Its source is not public yet. When it is, the quick start will look like this (from Mezze's README; you need Go 1.27.4 or newer, plus git and make):

git clone https://github.com/sengirunners/mezze && cd mezze && make install

cd ~/code/myapp
zm init                                    # starts the daemon and makes this repo a project
zm work new "Add the down migration"       # -> created myapp/local:k3x9qa
zm run --work local:k3x9qa claude          # the agent runs holding the item; detach with ctrl-]
zm send local:k3x9qa "the migration needs a down step"   # reaches whoever holds the item
zm ls                                      # sessions, the ones that need you first

make install puts zm, bd-zm, the drivers that lend a machine to a Potluck pool and potluck-conformance in Go's bin directory ($(go env GOBIN), or $(go env GOPATH)/bin), which you add to your PATH. zm doctor checks the machine's setup and says how to fix what is missing.

Mezze and Potluck

Mezze, the agent and team layer on Potluck, works on one machine with no pool. When your agents need more room, Potluck puts your team's idle Macs and Linux boxes to work as throwaway VMs and sandboxes, and the same make install puts its drivers on your machine so it can lend itself to a pool.

What Potluck does

Talk to Jackson and Clay, who make Mezze and Potluck

To use Mezze before its source is public, tell us what you run. We read every message.

Or write to us at ops@sengi.run.