ManageSelf-service
Developers run approved workflows from a form, ship one change across hundreds of repositories in a single review, and find any service and its owner in seconds. Your platform team keeps the guardrails on.
Service
Environment
From GitHub Environments
Version
Run status
Why it breaks at scale
Developers file tickets with the platform team for environments, secrets, deploys and pipeline changes, and wait.
With CodeCargo: Pre-approved workflows become forms developers run on demand, with the guardrails built in.
The same change across 50 repositories means 50 checkouts and 50 pull requests, so it takes weeks or never happens.
With CodeCargo: One developer ships a coordinated change across hundreds of repositories in a single review.
The architecture lives in tribal knowledge, service maps go stale the day they are written, and refactors break code nobody knew existed.
With CodeCargo: An AI service catalog keeps a live map of every service, owner, dependency and pipeline.
Self-service workflows
The platform team publishes vetted GitHub Actions once. Anyone with access runs them on demand, with the inputs, approvals and audit trail you set.
Compliant by default
Every self-service workflow runs on a building block your standards already check.
Pinned actions, least-privilege tokens and OIDC are part of the block, not something each team has to remember. Guardrails and CargoWall egress control apply to every run.
See how Secure works →Workspace · rename OrderClient.fetch to getOrder
50 repositories · one change · a pull request per repository
Multi-repository edits
Load every consumer of the API into a workspace, ask the AI agent for the change, review the diffs in one place, and ship a coordinated pull request per repository. No shell scripts, no spreadsheets, no tracking reviews by hand.
AI service catalog
CodeCargo derives services, owners, dependencies and CI/CD coverage from your repositories, so new hires understand the architecture in an afternoon and refactors stop breaking code nobody knew about.
How it works
Point CodeCargo at any workflow_dispatch action in GitHub. It becomes a building block, with no rewrite.
Set the inputs, who can run it and what needs approval, from the CodeCargo UI instead of YAML.
CodeCargo analyzes your repositories and keeps services, owners, dependencies and pipelines current.
Refactor an API, bump an action version or apply a security fix everywhere, in one review.
Part of Manage
Manage also keeps every repository on standard, fixes drift as pull requests, controls cost and cleans up hygiene, at the autonomy you set.
What platform teams ask before they open self-service to developers.
No. Self-service workflows are form-based: developers, and people who are not developers, fill in inputs and run. The YAML is written once by the platform team and reused.
Yes. Self-service workflows support three GitHub authorization modes: as the GitHub user, as a GitHub App for the whole organization, or as a GitHub App restricted to specific repositories. Access control is scoped to projects and the organization.
In GitHub. Self-service workflows run as GitHub Actions reusable workflows, so logs, artifacts and run history stay in the underlying repository.
A workspace loads the repositories you choose into one context for the AI agent. You review every diff in one place, then sync selected repositories back as pull requests, either new ones or updates to existing pull requests.
It is generated from your repositories in six stages: component discovery, ownership from Git history and CODEOWNERS, dependency analysis, CI/CD pattern detection, service and library boundaries, and a relationship graph. As code and ownership change, the catalog updates without manual upkeep.
CodeCargo never stores GitHub secrets; they stay in GitHub. Code is loaded into context for the AI editor only when a developer chooses to load it.
Yes. Self-service is part of CodeCargo Manage, alongside standards, security, cost and hygiene. Self-service workflows are built on the same building blocks your standards are checked against.