Skip to content

Getting started

Introduction

Learn how SuperPlane imports development tickets, runs coding agents and checks their changes.

SuperPlane is an open source factory for one-shot engineering. It turns high-confidence backlog issues into verified, review-ready pull requests.

SuperPlane coordinates coding agents, source control, CI, review, approvals and feedback in one workflow. Engineers can focus on work that needs judgment.

The factory finds routine work that agents can handle and applies consistent guardrails to every run. It returns failures with useful context.

Connect a repository and configure the Factory to listen for matching development tickets. The intake automation imports each ticket and records a confidence score from the checks configured by your team. Treat the score as a routing signal, not a guarantee: it helps separate bounded work with an observable outcome from work that still needs clarification.

When your team starts a task, SuperPlane dispatches the configured coding agent on a runner. The implementation automation records the run, branch and pull request. The verification automation then evaluates required status checks and pull request feedback. A failed check can return to the agent with its output attached; a passing result proceeds to human review.

For example, an issue that says “Add the YouTube channel to the Community page” can specify the target page, URL, accessible label and required new-tab behavior. SuperPlane can import that issue, run the change, check the link and build and present the resulting pull request with its run history. A person reviews the final diff and decides whether to merge it.

A repository issue enters Backlog, moves through agent implementation and automated verification and reaches the team with its pull request and check results attached.
  • Import and assess development tickets: Listen for matching work in connected trackers and record a confidence score before execution starts.
  • Run coding agents on configured compute: Start an agent with the repository, task context, model, tools and runner settings defined by the Factory.
  • Apply the same verification on every run: Evaluate CI status, repository checks and review feedback before the task reaches a person.
  • Connect your engineering stack: Build automations with more than 400 integration components across source control, CI/CD, observability, incident response, cloud infrastructure and AI providers.
  • Route failures with evidence: Attach command output, check results and pull request feedback to the next attempt instead of asking an engineer to reconstruct the failure.
  • Keep an execution record: Track task state, automation runs, artifacts, cost and the final outcome in a shared workspace.

The software factory adds a process around the agent

Section titled “The software factory adds a process around the agent”

A coding agent can inspect a repository, edit files and run commands. The software factory decides when that agent runs, which context and compute it receives, which checks must pass and when the result needs a person.

flowchart TD
  ticket[Development ticket] --> assess[Import and score]
  assess --> implement[Agent implements]
  implement --> verify[Automated verification]
  verify -->|Failed check with evidence| implement
  verify -->|Required checks pass| review[Human review]
  review --> outcome[Merge or close]

The normal path has one human review after the required automated checks pass. The Factory can pause earlier when a task needs a product, security or scope decision, but those exceptions are not additional review stages in the standard path.

Section titled “Example: add a YouTube link from a GitHub issue”

The GitHub issue asks for a YouTube logo and link in the Community section. The Factory records the request, its source and its expected result.

Start with a clear piece of work.

The example is small, but the process is reusable. Its implementation and verification rules live in Factory automations, where the team can inspect and update them for future tasks.

How a factory differs from an interactive coding agent

Section titled “How a factory differs from an interactive coding agent”

The distinction is not the model. A coding agent can run on a laptop or in a remote environment. SuperPlane adds the durable, team-level process around agent sessions.

Execution environment
Interactive coding agent sessionRuns on the developer’s computer or a remote environment selected for that session. The developer usually starts and supervises it.
SuperPlane FactoryDispatches each task to configured runner capacity, including on-demand managed compute, without requiring a developer’s workstation to stay online.
Instructions
Interactive coding agent sessionUses the prompt, checkout, tools and local configuration available in that session. Two developers can run the same task differently.
SuperPlane FactoryStores implementation and routing rules in shared automations so the team applies the same process across tasks and repositories.
Verification
Interactive coding agent sessionCan run tests and inspect output, but the commands and response to failure depend on the session instructions.
SuperPlane FactoryRuns the configured status checks, records their results and routes failures or review feedback back into the workflow before human review.
Continuity
Interactive coding agent sessionContext and progress are commonly tied to one conversation and one working environment.
SuperPlane FactoryPreserves task state, runs, artifacts, errors and pull requests in the workspace so another team member can inspect the same record.
Responsibility
Interactive coding agent sessionProduces the code change requested in the session.
SuperPlane FactoryCoordinates intake, agent execution, verification and escalation while leaving the final judgment with the team.

Read How SuperPlane works to learn how SuperPlane applies this model.

We would love to hear from you or help you get started. Use these channels to connect with the SuperPlane community.