For software teams
Every developer has a coding agent. Give your team a way to ship with them.
Volobox is where an agent's work becomes the team's work. Requests get sharp before anyone codes, checks and AI review run before a person is asked, and every pull request traces back to the customer who needed it.
We run our own company on Volobox. This is the path we built it for.
The problem
The code got faster. Everything around it stayed the same.
A coding agent helps one developer. Shipping is still a team's job, and this is where it gets stuck. Under each one: what changes in Volobox.
“Who asked for this, and why?”
An agent opens a pull request in one tool. The request, the discussion, and the decision behind it live in three others.
The pull request’s Story shows the ticket, the request, the task, and the session behind it.
“Review is where everything waits.”
Agents write faster than people can read. Pull requests queue up, and the first reviewer finds what a build would have caught.
Your build, test, and lint checks and an AI review run before a person is asked to review.
“CI is red and nobody is looking.”
The agent finished an hour ago. Its pull request has been failing since, and it stays that way until someone notices.
A CI repair workflow starts a repair round, and tells the team when it stops.
“Everyone has an agent. The team has none.”
Each developer runs their own agent on their own laptop. What it was told, what it did, and what it cost stay there.
A coding session is a shared room: the plan, the diffs, the terminal, and the cost of every turn.
“Where are we on this?”
The honest answer means reading git or interrupting someone. So status becomes a meeting.
Stories and Development status show what is moving, slowing, or stalled as it happens.
“Did the customer ever hear back?”
The fix merged last week. The ticket is still open, because support and the pull request never met.
A Close the loop card reaches the ticket’s owner when the linked work is done.
One piece of work
Follow one bug from the customer to the merge
A customer reports that captures made offline vanish when the app restarts. Here is every stop that report makes, and who does what at each one.
See this path on one of your own issues.
A free 30-minute demo, run on a task from your backlog.
01·Intake
It starts where the customer is
The customer’s ticket becomes a linked request. Issues and call notes start the same way.
Support & Portal02·Refinement
The request gets sharp before anyone codes
A workflow reads it with the code, asks what is unclear, and proposes acceptance criteria.
AI delivery templates03·Start
An agent takes the task, and you decide how far it goes
Plan mode waits for your approval; the session shows the diffs, terminal, and cost per turn.
AI Coding04·Checks
No pull request until your checks pass
Build, test, and lint must pass before it opens, with each result and the agent’s confidence.
05·Review
The review queue moves without being chased
AI review checks the diff. Workflows return findings to the agent, fix CI, and pick reviewers.
06·Merge
People decide what merges
Your approvals and branch protection decide. An agent’s approval never counts as a person’s.
07·Close the loop
The customer hears back, and the trail stays
A Close the loop card reaches the ticket’s owner, who replies. The path stays on the Story.
Story in Projects & Tasks
People stay in charge
The rules no workflow can change
The steps in between are workflows your team creates from the AI delivery templates, for the projects that need them. Each one is your own copy to read and change. These rules are not.
An agent never approves for a person
An agent’s approval does not count as a person’s, in any workflow, on any pull request.
Your branch protection is the floor
A workflow can ask for more than your code host requires before a merge. It can never ask for less.
AI review is advice
It reports findings and a verdict. It never approves a pull request and never blocks a merge.
Rounds are capped
Fix rounds and CI repair rounds are capped on every pull request, so an agent does not loop while nobody watches.
Every step leaves a note
The workflows write short notes in the task and pull request discussions, so the team follows the work without opening a workflow run.
Spend has a limit
Set limits for AI spend and sandbox compute. An administrator can turn AI off for the whole workspace or one area at once.
Your agents, your machines
Keep the coding agent your team already uses
Volobox does not replace your coding agent; it is the room the agent works in. OpenCode runs on your organization's models, Cursor and Codex on each person's own account, and Claude Code on a person's own subscription, on their own machine.
Sessions run in the cloud, on a developer's computer, or on a shared machine. The runner connects out, so no port opens, and shared work never runs on a personal computer.
Repositories stay on GitHub, GitLab.com, or Azure Repos, and a session can hand off to a second one. See AI Coding.
Shared context
Agents that already know the project
An agent in Volobox starts from the task, its discussion, and the request behind it, with the project's docs within reach.
Write the project's conventions once as agent guidance, and every agent follows them. What the team settles in a conversation becomes a learned note once a project admin approves it.
The AI reviewer reads the house rules in REVIEW.md from the target branch, so a pull request cannot rewrite its own rules. See workspace knowledge.
Visibility
See where delivery stands without asking anyone
Stories lists the work moving right now. Development status groups branches by person and marks each Moving, Slowing, Stalled, or In review. The Delivery card shows merges, the agent merge rate, and time to merge.
For anything else, describe the question and Create with AI draws the chart. An alert rule tells you when a number crosses its line.
Every coding turn shows its cost, administrators see usage under Reports, and Deployments follows your Azure DevOps pipelines. Explore Dashboards & Insights.
Beyond the pull request
The rest of the team's week is here too
Delivery is the sharpest example, not the whole product. The same workspace, and the same agents, carry the work around it.
Calls and call notes
Start a call from any chat. Call notes keep the decisions and action items, and an action item can become a task.
Learn moreAn agent in the call
Invite an agent by voice and ask it a task’s status or who owns something. It answers with its own access and never changes anything from a call.
Learn moreSupport & Portal
Customers open tickets in their own Portal, apart from the workspace. A ticket links to the request or task that fixes it.
Learn moreClient projects
Building for clients? Share chosen views of a project with them in the Portal, where they also open requests.
Learn moreDocs & Files
Specs, decisions, and runbooks next to the work, edited by several people at once. Agents answer from them with citations.
Learn moreA delivery team pack
Install the Software Delivery Team pack: a Release Manager for release notes and checklists, a Delivery Coach, and a Tech Writer.
Learn moreCustom Apps
Describe the internal tool your team is missing and publish it into the workspace, where it can react to a task closing or a pull request merging.
Learn moreAPI, MCP, and webhooks
Call Volobox from your own scripts and agents over its HTTP API or hosted MCP server, and receive signed webhooks when a pull request merges.
Learn morePer role
What changes for each role
Engineering leads
“Where are we, and what are the agents costing us?”
- Stories, Development status, and the Delivery card answer it without a status meeting
- Agents work under roles, approvals, budgets, and an audit log, like any other member of the team
- One process for every agent pull request, whichever coding agent a developer prefers
Developers
“I review agent output all day instead of building.”
- An agent’s pull request reaches you after build, test, and lint passed and the AI review’s findings were handled
- Request changes on a line and the agent takes the review. You are mentioned when it has pushed
- Your own coding agent, on your own machine if you want, in the same room as the team
QA
“I find out what changed after it is merged.”
- Every branch runs as a live app in a branch preview, with no local environment to maintain
- Acceptance criteria travel from the request to the task, and the AI review checks the diff against them
- The task’s Story shows what was asked, what was decided, and what changed
Product & support
“Is the customer’s fix shipped yet?”
- A ticket, its request, its task, and its pull request are one linked chain you can open from either end
- A Close the loop card reaches your Inbox when the linked work is done
- Ask an agent what happened with a ticket and it reads the same Story, with your access
Questions engineering teams ask first
Do we need to replace all our existing tools to use Volobox?
No. Most teams start with one team or one workflow, keep the rest of their stack, and connect it. GitHub and GitLab issues can mirror into projects, Google and Outlook calendars sync in, and MCP connections reach systems you already run. Adoption grows piece by piece; nothing forces a big-bang migration.
Full answerHow is this different from a standalone coding agent?
The coding room is part of the team workspace. The agent starts from a task and its context, while teammates can watch the terminal, diffs, plans, questions, and cost. The resulting branch, preview, pull request, and discussion stay connected to the work that started them, and the task’s Story shows how it travelled.
Full answerCan coding sessions run on our own machines?
Yes. A person can install a small runner on a computer they own, and their private sessions run there. Administrators can add shared machines and grant them to people or groups. The runner connects out to Volobox, so the machine opens no port, and it accepts only updates signed by Volosoft. Work shared with others never runs on someone’s personal machine.
Full answerHow is pricing and usage measured?
Per seat, from $24.90 a month at launch pricing, in three plans that differ in capacity, not features. You start with LLM and sandbox credit; usage after it is billed on your invoice.
Full answerIs there a free trial?
Yes. After a free demo, we set up a 30-day free trial of your own workspace on the plan you are evaluating.
Full answerGetting started
Start with one repository
Nothing here needs a migration. Keep your tracker and your code host, and add one piece at a time.
- 1
Bring one issue
A free 30-minute demo, run on a task from your own backlog instead of a sample.
- 2
Get your workspace
We set up your own cloud workspace the same day, with a 30-day free trial.
- 3
Connect one repository
GitHub, GitLab.com, or Azure Repos. GitHub and GitLab issues can mirror into the project.
- 4
Declare your checks
List build, test, and lint as verify steps in the preview manifest. /init-preview-manifest writes the first manifest for you.
- 5
Add workflows as trust grows
Create the delivery workflows the project needs one at a time, starting with AI review.
Bring one open issue to the demo.
We'll run it through the loop on the call: assignment, sandbox, checks, review, pull request.