Quick Start
Get started in 30 seconds. Download, install and launch your first AI session.
1. Download and install
Download PromptOps Manager from the Download page. Available for macOS, Windows and Linux. The installer takes care of everything — no configuration needed.
2. Sign in or register
Create an account or sign in with your credentials. Email verification uses a 6-digit OTP code. Your device is automatically registered for security.
3. Connect a workspace
Select your project directory. PromptOps automatically detects the structure, language, configuration files and database connections. The workspace is linked to your device.
4. Launch a session
Choose an AI provider (Claude, Codex, Gemini, Copilot or Shell), write a prompt and the agent starts working. Real-time output in the integrated terminal.
Tip: with Claude Pro/Max you get the full experience — sub-agents, extended thinking and up to 1M token context.
Sessions
Sessions are the heart of PromptOps Manager. Each session is an isolated environment with its own AI provider, terminal and context.
Creating a session
Click "New session" or use the shortcut. Select the desired AI provider and workspace. Each session has its own isolated terminal process via node-pty.
Multiple sessions
You can have multiple active sessions on the same workspace. Each session has its own terminal, context and independent state. Switch between sessions with the tabs in the top bar.
Lifecycle
Sessions are auto-saved every 30 seconds. In case of crash or accidental closure, the session is automatically restored on the next launch. You can also create reusable templates for frequent sessions.
Session history
Every action within a session is tracked: spawn, prompt, output, merge, errors. History is persistent on the backend and displayed as a git-style timeline.
Smart Context Injection
At session startup, the workspace is analyzed and a project map (structure, entry points, recent files) is injected into the agent context. Estimated saving: 30K-100K tokens per session.
Sub-Agents
Sub-agents are specialized AI agents that work in parallel under the supervision of the main agent.
Available roles
Each sub-agent has a predefined role with a dedicated system prompt:
- Security Audit — vulnerability analysis, code review for security flaws
- Test Runner — generation and execution of unit and integration tests
- Code Review — code review for quality, best practices and maintainability
- Documentation — generation and update of project documentation
- Refactoring — code restructuring to improve readability and performance
- Performance — bottleneck analysis and performance optimization
Spawn and management
One click spawns a sub-agent from the Quick Agents panel. Each sub-agent receives its own dedicated terminal, optimized context (with actually modified files from git diff) and reduced scope instructions.
Inter-agent communication
Agents communicate with each other automatically via the Agent Relay. The main agent modifies a file, the Security agent reviews it, the Test agent updates the tests. Asynchronous message passing with acknowledgment.
Result merging
Results from each sub-agent are collected and presented in the main session timeline. You can inspect each sub-agent's output individually.
AI Providers
PromptOps supports 5 AI providers, each with its own characteristics. You can switch providers per session or per individual sub-agent.
Claude Code
RecommendedThe deepest integration. Unlimited parallel sub-agents, extended thinking, up to 1M token context. Requires Anthropic Pro or Max plan. Recommended experience.
OpenAI Codex
Multi-file code generation with streaming output. Prompt library, session tracking and persistent history.
Gemini CLI
Large context codebase analysis. Session management, prompt versioning and team sharing.
GitHub Copilot
Inline completions and contextual suggestions. PR workflow integration and terminal command generation.
Shell / Custom CLI
Any CLI tool as a provider. Custom scripts, automation and full terminal access with integrated logging.
To switch providers, open session settings and select the desired provider. The switch is instant and does not interrupt the session.
Prompt Library
The Prompt Library is the structured prompt management system. Every prompt is a versioned, forkable and shareable asset.
Creating a prompt
Create a new prompt with title, content, category and tags. You can add dynamic variables with the {{variable}} syntax to make the prompt reusable across any project.
Versioning
Every change to a prompt creates a new version. You can compare versions with the diff viewer and restore a previous version at any time.
Fork and sharing
Fork a colleague's prompt to customize it. The derivation chain is automatically tracked. You can share prompts with the team — change requests go through an approval flow similar to a pull request.
AI Generator
Generate structured prompts from a task description. The generator automatically creates title, content, variables and template.
Built-in Git
PromptOps integrates a complete git system in the sidebar. No need to leave the app to manage versioning.
Supported operations
Stage, unstage, commit, diff, pull, push, stash, branch — all fundamental git operations are available from the UI. The Git Explorer shows the file tree with real-time status.
AI Commit & Branch
The system analyzes staged files, session context and linked task to generate commit messages in Conventional Commits format. Branch names are automatically generated from the task context.
Diff Viewer
View differences between files and commits with syntax highlighting. Compare versions side-by-side directly from the session.
Conflict resolution
In case of merge conflicts, AI assists in resolution by suggesting the most appropriate solution based on the project context.
Database Explorer
The Database Explorer automatically detects the connection from the project and allows browsing data in read-only mode.
Automatic connection
PromptOps analyzes the project looking for configuration files (.env, config/database.php, prisma/schema.prisma) and automatically detects the database connection. Supports MySQL, PostgreSQL and SQLite.
Data browsing
Browse tables, filter and sort data, inspect the schema. All operations are read-only for safety — you cannot accidentally modify the data.
Schema Inspector
View table structure: columns, types, primary keys and relationships. Useful for giving context to the AI agent when working with the database.
Voice Control
Send voice commands to the AI agent without touching the keyboard.
How it works
Press the microphone button (or the dedicated shortcut) and speak. The native macOS speech-to-text system transcribes your command into a prompt sent to the agent. Currently available on macOS only.
Tips
Speak clearly and concisely. The system works best with direct instructions. You can dictate complex prompts — the transcription handles punctuation and formatting.
Team & Collaboration
PromptOps Manager supports team-level collaboration with roles, permissions and structured sharing.
Creating a team
Create a team from the Team panel. Invite members via email — the invitation has a tracked status (pending, accepted, declined). Each member has a role: owner, admin or member.
Roles and permissions
Owner: full control, member and settings management. Admin: session, prompt and document management. Member: access to shared prompts and team sessions.
Sharing
Prompts, sessions and documents can be shared at the team level. The team lead sees all development timelines and workflows across the organization.
Documents and notes
Create contextual notes and documents linked to sessions, files or tasks. Documents are typed by category with full-text search and are reusable cross-project.
Token Efficiency Engine
The token optimization system reduces consumption and tracks savings in real time.
Smart Context Injection
At session startup, a project map (structure, entry points, recent files) is injected. The agent knows immediately where to go — no blind exploration. Saving: 30K-100K tokens per session.
Sub-Agent Prompt Optimization
Sub-agent prompts are enriched with actually modified files (from git diff) and reduced scope instructions. The agent reads 5 targeted files instead of exploring 30. Saving: 50K-70K tokens per sub-agent.
File Read Cache
Tracking of every file read with MD5 content hash. Redundant re-reads (unchanged files) are detected and flagged.
Dashboard
In the dashboard: total tokens used, tokens saved, average efficiency and session ranking by savings. Data aggregated from all sessions with database persistence.
Optimization Flags
Each session shows which optimizations are active (Smart Context, Cache, Graph, Tool Filtering, Local Execution) with an explanation of how each contributes to savings.
Docker & Integrations
PromptOps integrates with Docker and other tools in your development stack.
Docker Status
Monitor your project Docker containers directly from the app. See status and metadata without switching context. Monitoring is built into the workspace panel.
Hooks and automations
Configure hooks on session events: on_start, on_prompt, on_commit, on_end. Each hook runs a shell command. Pipelines are shareable at team level.
Code Editor
Open, edit and save files without leaving the session. Built-in editor with syntax highlighting and automatic language detection.
Projects & Workspaces
Workspaces link local directories to remote projects, maintaining consistency across multiple devices.
Workspaces
Each workspace associates a local directory with a remote workspace. The system uses the device hardware ID for synchronization — you can work on the same project from different machines.
Projects
Projects group workspaces, sessions, prompts and documents. Each project has its own members, roles and permissions. Sessions are automatically scoped to the active project.
Sprint Board
Integrated Kanban board with sprints (start/end date, velocity tracking). Tasks follow a state machine: backlog, in_progress, review, done. Many-to-many relationship between tasks and sessions to track which AI sessions worked on which task.
System Requirements
Minimum requirements to run PromptOps Manager Desktop.
macOS: macOS 12 Monterey or later. Intel or Apple Silicon chip. 4 GB RAM minimum.
Windows: Windows 10 (64-bit) or later. 4 GB RAM minimum.
Linux: Ubuntu 20.04+, Fedora 36+ or equivalent. 4 GB RAM minimum.
For AI providers: an active account with the chosen provider is required (Anthropic, OpenAI, Google, GitHub).
Internet connection required for AI provider communication and backend synchronization.
