Tasks the team can follow
Objectives, acceptance criteria, conversation, rounds, activity, Changes, Context used, and receipts stay together.
Collaborative AI software development
Let product, design, and engineering review AI work together.
Turn coding-agent output into shared product work the whole team can understand, inspect, and continue.
Iterel connects tasks, rounds, project context, Design, Changes, comments, and review without granting collaborators hidden access to a maker’s machine.
The team visibility gap
An agent transcript is not a shared product record.
Coding agents often work inside one person’s terminal, branch, or editor. Product owners and designers see the result late, while developers spend time translating prompts, screenshots, diffs, and decisions across separate tools.
Iterel makes the project legible outside the terminal. Work becomes a Task with rounds and receipts; informational questions become cited Chats; context, visual state, review, and comments remain connected to the product.
One shared loop
Collaboration around the work—not access to someone else’s machine.
Product context everyone can inspect
Uploads, Assets, Project Map specs, design-system state, task history, decisions, constraints, and accepted outcomes feed the shared record.
Roles with explicit authority
Signed-in members collaborate through project roles. Builder funding, permission, and machine authority remain separate decisions.
Design and review in context
Comments, presence, visual versions, task evidence, and exact Changes let different disciplines review the same product outcome.
Free review without a seat
One saved artifact, Design target, or Whiteboard target can be shared for View or Comment with unlimited reviewers and comments.
Attention across the project
Command center keeps running work, needs-input, failures, reviews, required actions, mission health, and the best next action visible.
Human-and-agent workflow
From request to a decision everyone can trace.
- 01
Frame the product work
Create the task with its objective, acceptance criteria, context, dependencies, and responsible people.
- 02
Run on the maker’s engine
The selected connected engine executes under the task’s local worktree, rules, and permission boundary.
- 03
Bring the evidence back
Activity, Changes, checks, visual evidence, Context used, and the round receipt return to the shared task.
- 04
Review in the right scope
Members work in the project; outside reviewers receive only an exact saved target with View or Comment permission.
- 05
Keep the decision
Accept, discard, or continue the task while its outcome and updated project context remain available to future work.
Trust boundary
Collaboration never silently becomes machine control.
Project role ≠ machine authority
A collaborator cannot run commands against another person’s Mac, credentials, or working tree merely because they joined the project.
Remote control is same-user
Eligible browser or phone control uses the same signed-in user’s online Mac; it is not a machine-off cloud coding promise.
Review is an exact projection
The public review shell exposes one saved target and comments—not Studio navigation, tasks, Context, Assets inventory, agents, tools, or settings.
Two kinds of sharing
Project access and Share for review solve different jobs.
Use project access for ongoing team collaboration. Use Share for review when someone should inspect one saved outcome without entering the Studio.
| Decision | Project access | Share for review |
|---|---|---|
| Scope | The full signed-in project, filtered by the member’s role. | Exactly one saved artifact, Design target, or Whiteboard target. |
| Audience | Invited project members. | Public viewers or commenters using the review link. |
| Authority | Role-based project collaboration; machine authority remains separate. | View or Comment only; no project, agent, tool, or machine authority. |
| Local project | Live collaboration requires an explicit collaborative cloud copy. | Publishes one immutable target snapshot and leaves the project local. |
| Free boundary | View members are free; maker access follows the plan and funding rules. | Review links, viewers, commenters, comments, and review collaboration are unlimited and Free. |
Questions, answered
Collaborative AI development without blurred authority.
Can product managers and designers review coding-agent work?
Yes. Project members can follow tasks, context, Design, Changes, activity, and comments according to their role, without needing access to the maker’s terminal.
Do reviewers need a paid seat?
No. View members and exact-target public viewers or commenters are free. Paid Builder access is a separate maker-authority and funding decision.
Can I share a whole task or live Preview publicly?
No. Public Share for review is deliberately limited to one saved artifact, Design target, or Whiteboard target. Tasks, Preview, project navigation, and machine controls remain private.
Can a collaborator control my local coding engine?
No. Project membership never grants another person authority over your Mac, credentials, commands, or working tree.
What happens when a local project needs live collaboration?
The owner explicitly creates a collaborative cloud copy. The local project remains, and the transfer excludes machine paths, credentials, raw logs, and local Asset bytes.
Bring the team into the loop
Make agent work reviewable before it becomes product history.
Free includes one editable collaborative cloud project and unlimited exact-target review links, viewers, commenters, and comments.