Skip to content
AnnouncementInitial release: self-hosted on Ubuntu 24.04 and 26.04 LTS. More platforms planned; Cloud and managed services are outside this release.First release: self-hosted on Ubuntu LTS.Release status
Register Cloud interest
Coding-agent orchestration software

Turn scoped engineering work into review-ready pull requests.

BranchRunner coordinates coding agents in isolated workspaces, applies validation and approval policies, and records evidence from issue to pull request. It is the operational layer around agents—not another coding model or IDE assistant.

Ubuntu 24.04 and 26.04 LTS. Source and release artifacts will be published when ready; more platforms are planned.
Inspect a sample run
Run 18attempt 2 of 2env-br-482-aruns/br-482
Task · TASK-482
Retry failed webhook deliveries with exponential backoff
R1Backoff is exponential with jitterSatisfied
R2Retries stop after five attemptsSatisfied
R3Alert fires on final failureSatisfied
R4Checkout flow unaffected in previewSatisfied
R5Queue schema unchangedVerified
Agent sessions
PlannerPlan approved · 4 requirements mappedCompleted
Implementersrc/workers · 48.2k tokens · $1.84Running
ReviewerWaits for validated handoffQueued
Validation and evidence
PassedUnit tests · 212 passed, 0 failed48s
PassedType check · no new errors21s
RunningPreview smoke test · env-br-482-a1m 04s
+41−7across 3 files · src/workers/webhook.ts
Run budget$4.20 of $12.00
Awaiting approvalApprove for independent review
Runs with the tools you already use
Coding agents
OpenAI Codex · Claude Code · Gemini CLI · OpenCode
Git and issue providers
GitHub available · GitLab, Linear and Jira planned
Runtime
Docker · Docker Compose · Git worktrees · no Kubernetes required

Coding agents are powerful. Operating them as a team is still difficult.

A single agent in an editor helps one person for one session. Running agents as part of an engineering organization exposes a different set of problems.

Agent sessions lose context

The objective, the decisions and the conversation live in a chat window that ends with the session.

Concurrent agents conflict

Two agents on one checkout overwrite each other. Nobody can tell which change belongs to which attempt.

Environments are inconsistent

Work that passes on a laptop fails in CI, and the environment that produced the result is already gone.

Human approvals are disconnected

Approval happens in a side channel, after the fact, without the diff and the test result in front of it.

What is BranchRunner?

BranchRunner is coding-agent orchestration software for engineering organizations. It coordinates agent work in isolated workspaces with validation, approval policies, and issue-to-pull-request evidence.

Who is it for?

It is designed for engineering, platform, development, and security reviewers evaluating how to operate coding-agent workflows with visible controls.

How is it different from a coding agent?

A coding agent performs work in a session. BranchRunner is the operational layer around agents, workspaces, checks, evidence, and human decisions.

Can we deploy it ourselves?

The self-managed operating model is documented, but a released artifact, supported-host matrix, and license terms are not published. It is not currently a deployment offer.

What should we do next?

Inspect the fictional sample run and pilot scorecard, then follow the initial self-hosted release. Cloud and Enterprise forms record future interest only.

How a task becomes a pull request

See the full lifecycle
  1. 01
    Define or import a task

    Write the objective in BranchRunner or import it from GitHub, Linear or Jira. The task holds the requirements and the human conversation for as long as the work lives.

  2. 02
    Provision an isolated workspace

    A run reserves a worker, creates a Git worktree from the local repository mirror and starts a Docker workspace with the services the task needs.

  3. 03
    Coordinate primary and specialist agents

    A primary agent plans the work and delegates to specialist sessions. Parallel work uses separate worktrees, then integrates in a controlled merge workspace.

  4. 04
    Validate and review every change

    Tests, type checks and preview checks attach to the requirement they satisfy. An independent reviewer reads the diff against the ledger, not the agent's summary.

What the platform does

Durable tasks and repeat attempts

The task keeps the objective, requirements and conversation. Each run is an attempt against it, so a failure is a retry rather than a restart.

Task orchestration
Multi-agent delegation you can follow

A primary agent delegates to specialist sessions. Every session shows its role, provider, current step, tool activity and cost while it runs.

Multi-agent execution
Isolated Git and Docker workspaces

Each run gets its own worktree and container set. Agents never execute on the control plane, and parallel work never shares a checkout.

Isolated workspaces
Human approvals, intervention and budgets

Redirect a run mid-flight, require approval before destructive tool use, and stop work at a spend, token or time limit.

Human governance

One task, end to end, in the product

TASK-482 in an example repository: the objective as it was written, the plan the agents worked from, what they did, what changed, what was verified, and the evidence a person approved.

Fictional example, not customer evidence

Inspect the sample objective, scope check, validation record, approval path, and a failure/recovery state without submitting contact details.

Inspect the sample run
Diagram of a fictional task moving from scoped requirements through an isolated run, validation, and human approval.
A crawlable diagram of the fictional workflow used in the sample run. The records and identities are examples, not customer evidence.

Interface built from the product model. Repository, task and identities are examples.

Choose where BranchRunner runs

Compare all editions
Community
Initial release · self-hosted

The first release targets Ubuntu 24.04 and 26.04 LTS. Source, artifacts, and complete licensing terms will be published when ready.

·Ubuntu 24.04 and 26.04 LTS; more platforms planned
·Minimum host: 2 CPU cores, 8 GB RAM and 60 GB free disk; 16 GB RAM and 4 cores recommended
·You own support, backups, retention, and deletion
Review Community requirements
Release validation is in progress. No support SLA or managed data services are included.
BranchRunner CloudFuture offering
Outside the initial release

Cloud is a future managed deployment offering. Only the self-hosted product is included in this release.

·Managed control plane and workers
·Best for teams that want convenience
·Upgrades, backups and capacity handled
Register Cloud interest
Cloud is outside the initial self-hosted release. Future availability and terms are not a current service offer.
Enterprise
Future offering · outside the initial release

Future organizational policy, identity, retention and deployment capabilities under consideration.

·SSO, advanced RBAC and exportable audit logs; SCIM planned
·Best for larger and regulated organizations
·Priority support and contractual commitments
Register Enterprise interest
Enterprise is a future offering; deployment modes and support are not yet available.

Give your coding agents a system of work.

Follow the self-hosted release for Ubuntu 24.04 and 26.04 LTS. Source and artifacts will be published when ready; managed offerings are outside this release.

Follow the self-hosted releaseReview Community requirementsRegister Enterprise interest
Cloud is outside the initial self-hosted release. Future availability and terms are not a current service offer.