Rally
Private beta

You. Your teammates. Your agents. Now in one place.

Rally is a shared workspace where you, your teammates, and your AI agents work side by side — in an editor panel or a web dashboard, backed by the same live project. Tasks, investigations, decisions, and agent jobs stay in one connected record, so nothing gets re-explained and no agent starts cold.

Private, early beta. Rolling access, not a waitlist that never moves.

checkout-service
Live
Feed
Tasks
Investigations
Decisions
Agent Jobs
  • Claude Code completed Agent Job #42

    Refactored auth middleware to use rotating session tokens · 3 files

    2m
  • P

    Priya dropped a steering message on Agent Job #42

    “skip the migration file, we'll handle that separately”

    6m
  • P

    Sam claimed Task

    Fix flaky checkout test on CI

    18m
  • CI Bot opened Investigation

    Why did staging deploy latency spike after last release?

    41m
  • P

    Priya recorded a Decision

    Use row-level security for tenant isolation, not app-layer checks

    1h
The problem

Work has to be re-explained constantly.

Not because people are careless — because nothing they use holds context long enough for it to still be there when they need it again.

Tasks live in one tool

Linear has the ticket. Slack has the reasoning behind it. Notion has a doc nobody linked back. None of them agree on what's actually true right now, so you end up asking a person instead.

Decisions get buried

The thread where you decided against a migration is three scrolls deep in a channel nobody re-reads. Six weeks later someone proposes the same migration, and nobody remembers why it was already rejected.

Agents forget everything

Every agent session starts from zero. It reads whatever files it opens, does the work, and forgets all of it the instant the session ends — including for the human who has to re-explain it to the next session.

How Rally works

One backend. Two surfaces read and write to it.

The Project Brain isn't a feature inside Rally — it's the thing Rally is. The editor panel and the web dashboard are both just views into the same live project, always in sync, because there's only one copy of the data.

Editor panel

Rally built directly into your editor. Claim a task, open an investigation, or review what an agent did overnight — without leaving your code.

Project Brain

The shared backend. Tasks, investigations, decisions, and agent jobs — one connected, searchable record, live for everyone reading it.

Web dashboard

The same live project, in a browser. No editor required — for planning, review, or checking in on a project from anywhere.

Agents connect to the Project Brain the same way — as actors on the project, not visitors. More on that below.

The comparison

Without a shared brain, versus with Rally.

Tasks · without

Scattered across Linear, Slack, and Notion — none of them in sync.

Tasks · with Rally

One connected, searchable record. Everyone reads from the same list.

Decisions · without

Buried in a Slack thread nobody will scroll back far enough to find.

Decisions · with Rally

Recorded once, in the project. Findable by anyone, at any point later.

Agent sessions · without

Every session starts from zero, and forgets everything the moment it ends.

Agent sessions · with Rally

An agent picks up a job with full context already loaded — no re-briefing.

Morning check-in · without

“What did the agent actually do overnight?” Silence, or a wall of scrollback.

Morning check-in · with Rally

You open the project and see exactly what happened, in order, as it happened.

Repeat work · without

Someone re-investigates something the team already answered last month.

Repeat work · with Rally

Rally flags the overlap before the new task or investigation gets created.

Feature breakdown

What the Project Brain actually holds.

This isn't another task tracker with an AI chatbot bolted on. Every feature below is a way of reading from or writing to the same shared record — for people and for agents.

Shared project memory

Tasks, investigations, decisions, and agent jobs live in one connected, searchable record. Both surfaces show the same live tabs — Feed, Tasks, Investigations, Decisions, Agent Jobs — real views into the data, not just forms to add more of it.

Agents as first-class participants

Connect Claude Code, a CI bot, or any MCP-compatible agent as a real actor on the project. It reads the same shared context other teammates do, and writes back into it — it doesn't start cold every session and vanish without a trace when it's done.

Steer agents mid-task

While an agent is actively working a claimed job, anyone on the project can drop in a steering message to redirect it. Course-correct without stopping it, taking it over, or losing its progress.

Live, not static

A real-time feed and presence indicators show what's happening right now — who's online, what they're working on, what an agent just did. Open a project and the state in front of you is current, not a snapshot from whenever someone last refreshed.

Duplicate-work detection

Before creating a task, investigation, or agent job, Rally checks for something that already covers the same ground and flags it. Effort doesn't get wasted re-discovering what's already known.

Real teams, GitHub-style membership

Proper accounts, invite-by-email, accept or decline, an owner role that manages who's in. No shared links, no guessing who has access.

Built with real security hygiene

Password reset, session management, rate-limited login, revocable access tokens. The account layer is built like it matters, not bolted on as an afterthought.

Built for teams and their agents

An agent is an actor on the project, not a bystander.

Most tools treat AI as a chat window bolted onto the side. Rally treats an agent the way it treats a teammate: it has an identity, it reads the same shared context everyone else does, and what it does gets written back to the record — visible, searchable, and still there tomorrow.

That cuts both ways. A human can steer an agent mid-task without stopping it or taking it over. An agent can flag that someone's about to redo work it already knows about. Multiplayer, for people and for the software working alongside them.

P

PriyaEngineer

Claimed the checkout latency investigation this morning. Left a decision on record before heading into a meeting.

Claude CodeAgent

Picked up an Agent Job overnight. Read the investigation Priya filed, had the context already, didn't ask her to re-explain it.

S

SamEngineering manager

Saw the agent was heading toward a migration the team already rejected. Dropped a steering message mid-task to redirect it.

CI BotAgent

Opened an investigation when a deploy regressed. Flagged as a duplicate of one Sam already had open — no wasted work.

Demo

A full walkthrough is coming soon.

We're putting together a real, unscripted walkthrough of the editor panel and dashboard in sync — not a polished sizzle reel. Until it's ready, requesting early access is the fastest way to see Rally running on an actual project.

Coming soon
Early access

Rally is in private beta.

Not publicly available yet. We're bringing on teams in small batches to keep the feedback loop tight. Tell us a bit about your setup and we'll reach out when there's room.

Role
Team size
Current tools for task tracking
AI agent usage today

Only email is required. Everything else helps us prioritize rollout — skip anything you'd rather not answer.