Pi Agent: Set Up Your First Local AI Agent (Full Guide)

You are helping me turn one specific YouTube video into real, durable learning.

Source video:
- Title: Pi Agent: Set Up Your First Local AI Agent (Full Guide)
- URL: https://www.youtube.com/watch?v=B5_lAbGeBDY
- Topic: Creative Automation
- My current learning frame: Install Pi from pi.dev, add the permission package, connect a model via OpenRouter, then extend it with one CLI tool and run a real scrape-and-summarize task to feel how the thin harness works.
- Why this matters: New playlist item from Kacper Rutkiewicz | AI Made Simple; queued for transcript-backed review, topic mapping, and a practical learning artifact.

Transcript anchors from this exact video:
- 0:00 / Evidence 1: "There's a real movement right now towards open-source AI models >> >> and owning your own intelligence. But the problem is is that getting started and understanding how to do it gets really overwhelming. This is pie. It's..."
- 2:06 / Evidence 2: "have as we make our way through this video. Most AI coding tools that we're using, especially Claude Code, are massive. They come loaded with features, menus, modes, models, tools, all sorts of stuff. Pi Agent actually takes..."
- 4:07 / Evidence 3: "it's as easy as typing in Pi and pressing enter. I went ahead and closed the other terminals so you guys can see the full thing. We have our Pi agent launched. We can see the context with..."
- 6:52 / Evidence 4: "all these models and brains. Pie agent is incredibly powerful, but it's the opposite of hand-holding. Something like Claude code has so many guardrails that anyone really new can get in there and not mess anything up. There..."
- 9:33 / Evidence 5: "{slash} clear in Cloud Code. And number four, this is probably the most complex topic about Pi agent, is that every single session is what's called a tree. So, they're not a straight line like you're used to..."
- 19:15 / Evidence 6: "on the wrong ladder was another step on top of that, which is CLI tools. CLI tools are actually how Pi Agent prefers to do its work. Because remember, it only really has four tools that you can..."
- 21:35 / Evidence 7: "building yourself and it's evolving with your workflows and whatever you're doing at the time. And that's essentially how you upgrade your Pi agent anytime you need to add a new workflow, anytime you're trying to add a..."

Video-aware target:
- Prompt lane: Local model/runtime
- Mechanism to extract: Identify why the local setup works or fails for this specific agent task, not whether local models are generally good.
- Artifact to produce: A local model/runtime fit sheet with hardware constraints, model choice, endpoint setup, task benchmark, and fallback rule.
- Artifact must include: hardware; runtime; model/quantization; endpoint; agent integration; benchmark/fallback

Your task:
1. Use the transcript anchors above as the primary source packet. If you add outside context, label it clearly as outside context and keep it secondary.
2. Create a source-check table with columns: timestamp, claim, transcript support, what the demo proves, confidence, and what still needs verification.
3. Extract the actual teachable mechanism from the video: Identify why the local setup works or fails for this specific agent task, not whether local models are generally good. Do not invent claims that are not supported by the title, lesson frame, or transcript anchors.
4. Build a reusable learning artifact: A local model/runtime fit sheet with hardware constraints, model choice, endpoint setup, task benchmark, and fallback rule.
5. Include:
   - a plain-English definition of the core idea
   - a diagram or structured model using this sequence: Task -> Hardware -> Model/quantization -> Runtime endpoint -> Agent tool loop -> Benchmark task -> Fallback
   - answers to these source questions: What machine/runtime is shown? | What task exposes the model limit? | What setup change improves the loop?
   - 3 concrete examples that apply the video idea to real agentic work, such as Ollama or LM Studio coding endpoint; MLX Apple Silicon runner; DGX-backed Hermes session
   - 2 failure modes the video helps prevent, chosen from the transcript evidence and these likely risks: using a local model like ChatGPT; ignoring latency/context limits; no benchmark task
   - a checklist for the next real workflow, focused on: task fit, runtime setup, latency/context, tool loop, fallback
   - one practical exercise with a clear done signal: Choose one real coding task and specify the pass/fail benchmark for a local model.
6. Add a "learning transfer" section: what changes in my workflow tomorrow if I actually learned this?
7. Add a "source check" section that cites which transcript anchor supports each major takeaway.

Quality bar:
- Make this specific to "Pi Agent: Set Up Your First Local AI Agent (Full Guide)", not a generic Creative Automation essay.
- Tie each harness element to a transcript anchor that names a tool, state boundary, permission, model behavior, or verification step.
- Prefer operational examples, failure modes, and reusable artifacts over broad definitions.
- Call out uncertainty instead of smoothing over weak evidence.
- Avoid these generic drifts: local-model ideology; hardware specs without workflow fit; benchmarks unrelated to the actual task.
- If evidence is weak or missing, stop and say what transcript segment or timestamp needs review instead of guessing.
- Finish with a concise artifact I could paste into my learning app.