Skip to content

isumitsoni/awesome-ai-pm

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 

Repository files navigation

Awesome AI PM Awesome

A curated, opinionated list of resources for Product Managers navigating the AI era.

Not everything — just the signal. Filtered by a practicing PM who builds AI products and uses these tools daily.

Maintained by Sumit Soni · LinkedIn

Last updated: 2026-03-17


Contents


🧠 AI PM Foundations

Understanding what it means to be a PM in the AI era — not just using AI tools, but owning AI features end-to-end.


🛠️ Vibe Coding & Agentic Tools

The tools PMs are using to ship without a dedicated engineering team.

CLI Agents (for building end-to-end)

  • Claude Code — Anthropic's CLI agent. Best for structured SDLC, serious builds, and repo-level context. Skills + hooks + MCP support.
  • OpenAI Codex — OpenAI's coding agent. Good for agentic experiments and broad exploration.
  • Gemini CLI — Google's CLI. Best for large-context analysis, research, and multimodal tasks.

UI Prototyping (ship the interface first)

  • v0.dev — Vercel's AI UI generator. Describe → get production-ready React components.
  • Bolt.new — Full-stack prototyping in the browser. Clone → deploy in minutes.

Code Editors with AI

  • Cursor — AI-native code editor. Best for longer coding sessions with full codebase context.
  • GitHub Copilot — In-IDE pair programmer. Good entry point if you're already in VS Code.

Agentic Workflow Tools

  • BloopAI/vibe-kanban ⭐ 23k — PM-friendly workflow layer for coding agents. Helps non-engineers structure agent tasks and track progress.
  • OpenHands/OpenHands ⭐ 69k — Autonomous coding agent with broad community support and active maintenance.
  • Aider-AI/aider ⭐ 42k — Terminal-based coding agent built for iterative product build loops.
  • stackblitz-labs/bolt.diy ⭐ 19k — Open-source browser-based full-stack vibe coding. Strong for rapid founder prototyping.
  • PrefectHQ/fastmcp ⭐ 23.7k — FastMCP framework for exposing tools and workflows to agents. Useful reference now that MCP is showing up in more founder and PM build conversations.

PM-Specific Skill Libraries

  • phuryn/pm-skills ⭐ 7.3k — 100+ agentic skills for Claude Code, Codex, Gemini CLI, Cursor, and Kiro. Covers discovery → strategy → GTM.

💬 Prompt Libraries for PMs

Structured prompts for real PM workflows — not generic "write me a PRD" prompts.

  • isumitsoni/pm-prompts — Practical prompt library for every PM workflow: discovery, strategy, execution, AI feature specs, metrics analysis, and career. 8 categories, 53 prompts.
  • Claude Code Skills Marketplace — Slash commands and skills for Claude Code that automate PM workflows.

Know a good prompt library? Open a PR.


📊 Evals & AI Quality

How to define "good" for an AI feature — the PM's job, not the ML engineer's.

  • Anthropic's Guide to Evals — How to build test suites for LLM applications. The starting point.
  • AI Evals Course on Maven — Practical eval design for PMs. Highly recommended.
  • promptfoo/promptfoo ⭐ 16k — Evals + red-teaming framework PMs can use as a release gate. Runs prompt comparisons and catches regressions.
  • confident-ai/deepeval ⭐ 14k — Straightforward framework for building LLM evaluation suites. Good starting point for PM-led eval programs.
  • Arize-ai/phoenix ⭐ 8.8k — Observability + eval workflows for production AI features. Useful for ongoing quality monitoring.
  • truera/trulens ⭐ 3.1k — Feedback and eval instrumentation for LLM systems.
  • langfuse/langfuse ⭐ 23.3k — Open-source observability, evals, prompt management, and datasets. Good PM pick because it makes quality review and regression tracking visible beyond raw logs.
  • Helicone/helicone ⭐ 5.3k — LLM observability plus experimentation. Useful for PMs who want to compare prompts, models, and live behavior without building custom dashboards first.
  • tensorzero/tensorzero ⭐ 11.1k — Gateway plus observability, optimization, and evaluation in one stack. Strong addition for PMs thinking about production quality and experimentation together.
  • vibrantlabsai/ragas ⭐ 13.0k — Evaluation toolkit for LLM and RAG systems. Helpful when PMs need a more structured quality conversation than "the outputs seem better."
  • guardrails-ai/guardrails ⭐ 6.5k — Guardrail layer for LLM applications. Good PM resource for trust-sensitive workflows where failure handling matters as much as model performance.
  • comet-ml/opik ⭐ 18.3k — Open-source AI observability, evaluation, and optimization. Strong PM pick when you want traces, evals, and prompt iteration in one system instead of separate tools.
  • Giskard-AI/giskard-oss ⭐ 5.1k — AI testing framework focused on performance, bias, and security risk. Useful when PMs need a more explicit trust and red-team conversation before rollout.
  • Agenta-AI/agenta ⭐ 3.9k — Open-source LLMOps stack for prompt management, evaluation, and observability. Good fit for teams graduating from ad hoc prompt tests into repeatable experimentation.
  • Key PM mental model: Define your eval criteria before you build. "The AI should sound like a human" is not an eval. "95% of outputs score ≥4/5 on our rubric" is.

🏗️ Building AI Products

Patterns, frameworks, and hard-won lessons from PMs who ship AI features.

  • Vibe Coding vs. Vibe Engineering — Important distinction: vibe coding is prototyping; vibe engineering is production. Know which one you're doing.
  • The PM's Guide to RAG — When to use RAG vs fine-tuning vs prompt engineering, explained for PMs.
  • Shubhamsaboo/awesome-llm-apps — Working code examples for LLM applications. Clone → run. Great for PMs who want to understand by doing.
  • humanlayer/12-factor-agents ⭐ 18.8k — Production principles for LLM-powered software. Strong fit for PMs and founders because it sharpens architecture judgment without feeling like an ML textbook.

Key architectural decisions PMs own

Decision What to consider
RAG vs fine-tuning RAG for dynamic/private data; fine-tuning for consistent tone/style
Model choice Cost/latency/capability triangle — don't default to the biggest model
Streaming vs batch User experience decision first, then engineering
Memory architecture Session vs persistent vs none — depends on your trust model

💰 LLM Economics for PMs

Cost and latency are product decisions. PMs need to own them.

  • BerriAI/litellm ⭐ 39.3k — Model gateway with cost tracking, routing, logging, and guardrails. Very relevant for PMs making cost-versus-quality decisions across providers.
  • Portkey-AI/gateway ⭐ 10.9k — AI gateway with routing and integrated guardrails. Useful for PMs who need reliability, fallback logic, and vendor flexibility built into product decisions.
  • Token pricing reference: Anthropic pricing · OpenAI pricing · Google pricing
  • Mental model: Input tokens are cheap. Output tokens cost 3-5x more. Design your prompts accordingly.
  • Haiku/Flash for high-volume, low-stakes tasks. Sonnet/Pro for core reasoning. Opus/Ultra for one-shot complex analysis.
  • Caching: If your system prompt is long and reused, prompt caching can cut costs 80-90%.

📚 Learning Resources

Courses and programs worth your time.

Resource What it covers Who it's for
AI PM Certification (Maven) Full AI PM curriculum PMs transitioning to AI PM
AI Evals (Maven) How to evaluate LLM outputs PMs owning AI quality
fast.ai Practical deep learning PMs who want technical depth
Anthropic Claude Code Docs Building with Claude Code PMs who vibe code
microsoft/generative-ai-for-beginners ⭐ 108k Structured lesson path for AI builders PMs new to AI development
openai/openai-cookbook ⭐ 72k Implementation patterns and examples PMs prototyping AI features
anthropics/claude-cookbooks ⭐ 35k Practical Claude workflows and examples PMs building with Claude API

📡 Newsletters & Communities

High signal-to-noise sources on AI PM.

  • The Product Compass — Paweł Huryn. Best AI PM newsletter. Actionable, not hype.
  • Unwind AI — Shubham Saboo. Open-source AI builder community. High technical signal.
  • Lenny's Newsletter — Not AI-specific, but the best general PM source.

🔧 Open Source PM Tools

Tools PMs can use, fork, or learn from.


Contributing

This list is maintained by a practicing PM — not a curator scraping the internet.

  • Open a PR if you have a resource that genuinely belongs here
  • Quality bar: would you recommend this to a senior PM who has 30 minutes? If yes, submit it.
  • No self-promotion without real value. No courses you haven't taken.

Star this repo if you find it useful. It helps other PMs find it.

About

A curated, opinionated list of resources for Product Managers navigating the AI era — tools, evals, learning resources, and practical guides. Updated by a practicing AI PM.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors