Home
Open source · September 2026

Keep your team and AI tools aligned.

Backplane keeps work, decisions, dependencies, and review results on one shared board. Every handoff starts from the same project record, not a fresh chat.

Get the launch link See how it works

Workspace

Works without AI
Search cards…
Filter
Backlog2

Rate-limit the export API

featureVLR-184

Audit log retention

taskVLR-191
In progress1

Refactor auth middleware

taskVLR-176
Review1

Fix flaky upload test

bugVLR-170
Done1

Signed URL rotation

featureVLR-165

Used on real projects

We built Backplane with Backplane.

We used it first on our own projects. Then, to improve Backplane itself.

Figures are estimates across projects. The full history of each run lives on its Backplane board.

Backplane itself started as the internal tool of valaris.studio.

25+

Real projects

projects run with Backplane as the only project and context platform ~ from end-to-end products to specialized customer work like security hardening and reverse engineering.

7

Autonomous runs

projects delivered end to end by runner pipelines and loops, without step-by-step supervision.

~200

Cards per run

cards moved to done in a typical autonomous run, with thousands of lines of shipped code behind them.

The problem

The hard part is not assigning a task. It is keeping the project aligned.

People join, AI sessions restart, and decisions change. Without a shared record, the team repeats information, overlaps work, and loses sight of what is ready for review.

Decisions get lost

Why a path was rejected or which constraint matters should not disappear when a session ends.

Work collides

When ownership is unclear, two contributors can solve the same problem or undo each other's changes.

A status is not a review

Done only becomes useful when the team can see the result, the checks that ran, and the reviewer's decision.

Keep the project record where the work happens.

Backplane keeps cards, decisions, dependencies, and review results together. People and AI tools update the same board, so the next handoff begins with what the team already knows.

What a team needs

Six essentials for better teamwork.

One place for the work, its context, and the decisions that move it forward.

Shared work and state

See what is being done, who owns it, and what changed.

Context that does not disappear

Keep decisions, notes, and files beside the work they explain.

Tools that stay in sync

Connected tools read and update the same board as the team.

Delegation with clear rules

Every task has an owner, a goal, and clear limits.

Visible progress and costs

See what happened, what it cost, and when a limit was reached.

Human approval

Important actions and completed work always return to a person for review.

Six things a team needs. Three layers that deliver them.

See every screen

Layer 1 · Platform

Keep the whole project in one place.

Plan, share what the team already knows, and track progress. It works just as well before you connect any AI tool.

Search cards…
Filter
Backlog2

Rate-limit the export API

featureVLR-184

Audit log retention

taskVLR-191
In progress1

Refactor auth middleware

taskVLR-176
Review1

Fix flaky upload test

bugVLR-170
Done1

Signed URL rotation

featureVLR-165
Visible workOrganize tasks, owners, priorities, and dependencies on Kanban boards.
Decisions and filesKeep notes, decisions, and files beside the work inside a private workspace.
Clear objectivesSet scope, milestones, owners, and constraints before work begins.
Search cards…
Filter
Backlog2

Rate-limit the export API

featureVLR-184

Audit log retention

taskVLR-191
In progress1

Refactor auth middleware

taskVLR-176
Review1

Fix flaky upload test

bugVLR-170
Done1

Signed URL rotation

featureVLR-165
See every screen
Experimental

Execution loops

Keep repeated work moving within clear limits.

You define what each iteration should achieve. The runner rereads those instructions before every cycle and carries out the next step within the limits you set.

Set the maximum spend

The loop has a total spending limit. Resetting it requires an explicit decision by the person operating it.

Limit the number of attempts

The loop stops when it reaches the configured maximum, even if work remains.

Stop after repeated failures

Consecutive errors, empty results, or timeouts stop the run before more budget is spent.

Hand control back when needed

The instructions can stop the loop when the goal is met or a person needs to decide.

loop.yaml
loop_prompt: |
  Take the highest-priority card that is ready.
  Implement it and open a pull request.
  Stop the loop when no cards remain.

budget_usd: 20               # hard spending cap
max_iterations: 25           # attempt limit
max_consecutive_failures: 3  # stop after repeated failures
starvation_policy: park      # hand control back

Backplane reads the configuration at the start of each iteration. Review the log and update the instructions; the next cycle uses the change without a restart or redeploy.

We still label loops experimental ~ and they have already run successfully on more than 10 real projects, including Backplane itself.

See every screen
New

Skills

Save what works as a reusable skill.

Create a skill, assign it to a board, and the whole team follows the same process.

propose
approve
assign
prepare
run
All your skills in one place

Start from the catalog or create your own. Publish, assign, and archive from the same workspace.

Simple files you can take with you

Each skill is stored as Markdown, so your team can read, edit, and move it whenever needed.

The right skill for each board

Use the latest version or keep a specific one. Everyone works from the same instructions.

People review every change

Tools can suggest improvements, but nothing is shared until someone reviews and approves it.

Ready before work begins

Backplane prepares approved skills so Claude Code or Codex can use them from the start.

Archive, never delete

Remove old skills from active use without breaking boards or losing past versions.

Keep what works. Reuse it on the next project.

See every screen

Control

A great configurability comes with great responsibility.

Adjust each stage and write precise instructions. Backplane applies limits and checks so a poor configuration does not waste budget or send the work off course.

Adjust each stage and role
Rules applied by Backplane
Limits and automatic cutoffs
See every screen

Costs and activity

See what each run cost and what changed.

Review spending by run, role, card, and model. Set limits so Backplane can alert, pause, or stop work before it exceeds the budget.

Spent this cycle$54.40
orchestrator$38.4 of $60
reviewer$11.2 of $40
documentator$4.8 of $25

Cost breakdown

Per execution, role, card, and model.

Budgets

Limits by runner and workspace, with an automatic cutoff.

Budgets by role and project

Complete history

Review every run, tool used, and card changed.

See every screen

How it works

See the workspace before any AI tool connects.

A real tour of boards, cards, project definitions, and notes, captured directly from the product.

Design choice

Use the AI tools your team already trusts.

Backplane does not include or resell an AI model. Before you connect one, it still works as a fast, self-hosted project board.

Keep provider keys under your control

Backplane does not ask for AI provider keys. The usage and costs it records remain tied to your own accounts.

Bring the coding tools you already use

Runners currently execute Claude Code or Codex CLI through your own subscriptions, keys, and rules.

A coordination layer, not an AI provider

Backplane provides the shared workspace, rules, and history. The tools you connect provide the model capabilities.

Models change quickly. Your project record should remain stable.

Where Backplane comes from

It started as our own internal tool.

When coding agents turned familiar ground into unknown terrain, the valaris.studio team needed collaboration, trackability, and observability in their own software projects ~ workflows that held up while everything else changed.

That work distilled into a domain-free platform, and it has managed all kinds of projects since ~ ours and other people's.

Open-sourcing it is how we give back: helping others land their visions in the best way we know.

"Proudly agentic". At valaris.studio we've been able to deal with the swiss-cheese problem of working with agents. Every mid-session discovery, deferred task, and gap that well-intentioned agents leave behind is suddenly covered. And that's just one benefit of using Backplane as a contextual habitat for people and agents.
Sebastian Breit · Founder, valaris.studio

Choose how to run it

Choose who operates Backplane.

Run the open-source product on infrastructure you control. Backplane Cloud is planned for teams that would rather leave deployment and operations to us.

Self-hosted

Open Source

Run the complete product on your infrastructure, keep control of your data, and inspect or extend the source.

  • Complete platform, MCP server, and runner
  • Your infrastructure, your data
  • Extensible and inspectable
Planned

Valaris Backplane Cloud

We plan to host and operate the same platform for teams that prefer to focus on their work instead of maintaining infrastructure.

  • Managed platform and runners
  • Support and guided onboarding
  • Capacity that grows with your team
Tell me when it is available

Enterprise

Planned

Enterprise is planned too. It will include centralized identity management, Backplane keys, and organization budgets. It will not manage AI provider keys.

Talk to us

Frequently asked questions

Answers to the most common questions.

It is a self-hosted workspace where the team and its AI tools share the same Kanban board. It is not a chatbot, a copilot, or an agent framework. It also includes no AI model.

Pre-register

Get the repository when it opens.

Leave your email. At launch, we will send the repository, setup guide, and launch note. We will use the list for nothing else.