prjct/cli

Product Introduction

1. prjct/cli is a command-line interface tool that serves as an execution layer for AI development environments including Claude Code, OpenAI Codex, and Warp. It converts unstructured ideas into technical roadmaps, generates AI-ready task specifications, and enables direct feature shipping from the terminal without traditional project management overhead.
2. The core value lies in its ability to eliminate friction between conceptualization and execution by automatically adapting outputs to specific AI agents, maintaining persistent technical context across sessions, and providing a zero-ceremony workflow for solo developers and AI-assisted teams.

Main Features

1. Smart Agent Detection automatically identifies the active AI environment (Claude Code, OpenAI Codex, or Terminal/CLI) through environment variables, configuration files, and runtime analysis, ensuring optimal output formatting without manual configuration. Detection checks include CLAUDE_AGENT/CODEX_AGENT variables, MCP server availability, sandbox directory structures, and TTY session verification.
2. Context-Aware Roadmapping transforms natural language ideas into technical tasks using /p: commands like /p:idea and /p:roadmap, preserving architecture decisions, tech stack details, and progress history across sessions through local .prjct configuration files. Tasks maintain dependencies, estimated timelines, and agent-specific formatting requirements.
3. Execution Pipeline Integration enables direct terminal interaction through commands like /p:now for task focus and /p:ship for deployment, with ANSI color support, progress spinners, and automatic GitHub Actions/Codespaces compatibility. The system logs shipped features with timestamps and velocity metrics in machine-readable JSON format.

Problems Solved

1. Eliminates context fragmentation between AI agents and human developers by maintaining persistent technical specifications, architecture decisions, and progress tracking in a local .prjct directory that both parties can reference.
2. Targets indie hackers, solopreneurs, and AI-assisted development teams who require production-grade task breakdowns without Jira-style ticket systems or Scrum ceremonies. Particularly effective for developers shipping MVPs within tight timelines.
3. Addresses use cases including rapid conversion of feature ideas into Codex-compatible prompts, resumption of complex projects after context gaps, and generation of agent-specific instructions for authentication systems or API integrations.

Unique Advantages

1. Differs from traditional PM tools by operating as a CLI-first layer that modifies development environments rather than requiring separate platforms, using lightweight markdown-based task definitions (.prjct/roadmap.md) instead of complex ticket systems.
2. Introduces runtime environment fingerprinting that combines static analysis (CLAUDE.md/AGENTS.md checks) with dynamic verification (MCP server pings, sandbox path validation) to optimize outputs for each agent's capabilities.
3. Provides competitive advantages through local-first architecture with zero cloud dependencies, instant installation via curl -fsSL https://www.prjct.app/install.sh, and compatibility with existing AI workflows through standardized /p: commands.

Frequently Asked Questions (FAQ)

1. How does prjct/cli detect different AI agents? The tool performs four-layer detection: environment variable checks (CLAUDE_AGENT/CODEX_AGENT), configuration file scans (CLAUDE.md/AGENTS.md), runtime capability tests (MCP API availability), and directory structure analysis (/.claude or /sandbox paths).
2. Is project data stored in the cloud? No, all context including technical specifications and progress tracking remains local within the .prjct directory, with optional GitHub integration through /p:ship commands that push only user-approved artifacts.
3. Can I integrate prjct/cli with existing projects? Yes, run /p:init in any project root to create a .prjct config, which automatically ingests existing README.md and package.json files to establish baseline context for AI agents.
4. What happens when switching between Claude Code and Terminal modes? The CLI automatically regenerates task formats using Claude-specific markdown templates or Terminal-optimized ANSI outputs based on real-time environment checks.
5. How does progress tracking work? The system logs /p:done commands with timestamped completions in .prjct/progress.json, calculating velocity metrics through weighted task complexity scores visible via /p:progress.