Key Concepts

Understanding these core concepts will help you get the most out of proxifai.

Organizations

An organization is the top-level unit. It represents your team or company. Each organization has its own members, teams, services, repositories, billing, and settings. Members can belong to multiple organizations and switch between them.

Teams

Teams group members within an organization. Services, sprints, and views can be scoped to a team, so each team sees only the work relevant to them. A team also owns an identifier prefix used to number issues (e.g. BACK-42).

Services

A service is the unit you build, ship, and operate — an application, API, or component with its own code, deployments, and runtime. The service is also a scope filter: pick one from the top-bar scope picker and every page (issues, deployments, logs, alerts, cost) narrows to it. Picking nothing shows the whole workspace.

Under the hood a service is backed by a project record, which carries planning metadata: a status (backlog, planned, in_progress, paused, completed, cancelled), an independent health (on_track, at_risk, off_track, no_updates), an owner, linked repositories, and catalog fields (tier, lifecycle, language, tags, links). You will see both terms — “service” in the product UI, “project” in the API and data model — referring to the same thing.

Environments

An environment is a named deployment target for a service — dev, qa, preprod, prod, or an auto-expiring preview. Each environment points at a deployments (GitOps) repository and branch, can require approvals before changes apply, and can promote from a parent environment. See Environments.

Deployments

A deployment ships a service’s manifests to an environment’s Kubernetes namespace. proxifai is GitOps-native: a commit to the deployments repo (or a UI/CLI action that writes one) is reconciled and applied. See Deploy.

Issues

Issues are the fundamental unit of work. They have:

• Status — backlog, todo, in progress, in review, done, cancelled
• Priority — urgent, high, medium, low, none
• Type — task, bug, story, epic, subtask
• Assignees — humans or AI agents
• Labels, time entries, and threaded comments

Issues connect to branches and pull requests, creating full traceability from planning to code.

Repositories

proxifai includes a native Git forge. You host repositories directly on the platform — no external Git provider required. Repositories support branches, tags, releases, deploy keys, branch protection, and an integrated OCI container registry.

Pull Requests

Pull requests enable code review within the platform: diff viewing, inline comments, review verdicts, branch protection, CI/CD checks, and merge methods.

Sprints, Roadmaps & Initiatives

Sprints are time-boxed iterations scoped to a team. Initiatives are longer-range groupings that link issues across services and render on the roadmap timeline.

Workflows

Workflows are container-isolated automation graphs triggered by events, schedules, or manually. They cover both CI/CD workflows and general automation, and support conditional logic, loops, parallelism, and human-in-the-loop approvals. Workflow steps can dispatch AI agents that execute in isolated containers. See Workflows.

Agents

AI Agents run in fully isolated containers (Docker, or Kubernetes on Enterprise). They can:

• Clone repositories and install dependencies
• Write code, run tests, and open pull requests
• Triage issues and suggest priorities
• Answer questions about your entire codebase

Agents access only what their owner can access — no privilege escalation.

Knowledge Base & Models

The Knowledge Base indexes your code, documents, issues, and PRs for AI-powered retrieval. Use the assistant with @ mentions to reference anything across your organization.

The Models product is proxifai’s LLM gateway: one OpenAI-compatible endpoint that fans out to OpenAI, Anthropic, and Google, with bring-your-own-keys, failover, usage tracking, and budgets. See Models.

Operate: Insights, Alerts & Cost

Once a service is running, the Operate surfaces tell you how it’s doing:

• Insights — logs, metrics, and traces (LogQL/PromQL), plus user-built dashboards and an audit log.
• Alerts — on-call incidents (firing → acknowledged → resolved) and the alert rules that raise them.
• Cost — unified FinOps across compute, LLM, storage, seats, and network spend, with allocation, budgets, and optimization.

Inbox

The Inbox is a unified notification center with real-time updates. Notifications carry inline actions — approve a workflow run, merge a PR, or respond to a workflow approval without leaving the inbox.