Example page made by Ghostlaunch. This product doesn't exist yet.
Back to the exampleFor parallel Claude Code sessions
Agentry is a CLI and local dashboard for devs running several Claude Code sessions on one repo. It flags who needs you, claims files, and pushes decisions live.
Runs locally. Your code never leaves your machine.
The hard part is everything between the sessions: who's waiting, who's writing, and who's still working from this morning's assumptions.
"By lunch I can't tell which agent is blocked on me and which is still writing."
"Whoever saved last wiped the others. No errors anywhere."
"I spend half my day copy-pasting between windows."
"Agent B is still working off the schema it read at 9am."
"One agent sat idle for 40 minutes on a yes/no question."
Agentry sits between your Claude Code sessions and your repo, so you see them in one place and they stop stepping on each other.
Run agentry init in the repo your agents share. It sets up a local coordination store for that project. Nothing is uploaded anywhere.
Launch each Claude Code session with agentry run and a short label like auth-refactor. Agentry now tracks its state and the files it touches.
One view on localhost shows every session as blocked on you, working, or queued, with the question it's waiting on and how long it's waited.
Log a choice like "use httpOnly cookies for tokens" and Agentry pushes it to every running session, so none keeps acting on a stale snapshot.
Worktrees and tmux keep agents apart. Agentry keeps them in sync and keeps you in the loop.
Every session is tagged blocked, working or queued, sorted by how long it has waited. The question it's stuck on is shown right there, so you can answer without hunting for the tab.
Before an agent writes, it claims the file. If another session holds the claim, the second agent waits instead of saving over it, and you see the queue on the dashboard.
Decisions you log, or that an agent records, are delivered to all running sessions mid-task. No more agent B building on the schema agent A changed an hour ago.
Instead of copy-pasting one agent's conclusion into another's window, record it once in Agentry. Every relevant session gets it, and you get your afternoon back.
Agentry is for you if
You run 3+ Claude Code sessions at once
Separate tasks on one repo, all going at the same time, and you've lost track of which one needs you.
Your agents share a single checkout
You've had a file silently clobbered, or you'd rather not juggle a worktree per agent.
Your agents' decisions affect each other
Schemas, auth, API shapes: choices in one session that the others need to know about right now.
Probably not for you if
They isolate agents. Agentry coordinates them.
| Criteria | Agentry | Git worktrees | tmux / tabs |
|---|---|---|---|
| Shows which session is blocked on you | Yes | No | Cycle through every pane |
| Prevents silent file overwrites | Yes | Until merge time | No |
| Shares decisions between running agents | Yes | No | No |
| Updates context mid-session | Yes | No | No |
| Works on one shared checkout | Yes | One checkout per agent | Yes |
Start free with the status view. Upgrade when you want agents to stop colliding.
See every session's status in one place.
Free
Full coordination for devs running 3+ agents on one repo.
$19/ month
Everything runs on your machine on both plans. Cancel anytime.
Claude Code only at launch. Support for other agents like Codex and Cursor is planned, but we'd rather do one well first. Tell us which you use when you sign up.
No. Agentry is a CLI plus a dashboard served on localhost. Session state, file claims and decisions are stored locally. Nothing is sent to us or anyone else.
Worktrees keep files apart, but they don't tell you which agent is waiting on you, and they do nothing about two agents making conflicting decisions or working from stale context. You only find out at merge time.
The first session claims it. The second waits until the claim is released instead of overwriting, and it shows up as queued on the dashboard so you can step in if needed.
When a decision is recorded, Agentry delivers it to each running session it manages, so the agent picks it up mid-task rather than relying only on what it read at startup.
A few minutes: install the CLI, run agentry init in your repo, and start your sessions with agentry run. Then open the dashboard in your browser.
One silent overwrite or one agent idling for 40 minutes on a forgotten question costs more than that. The free plan lets you try the status view first.
Get early access to Agentry and run parallel Claude Code sessions without being the message bus.