Pie — deploy and operate your own AI agent
Provision a personal autonomous agent on free cloud infrastructure, run your daily work through it, and get coached on how you operate it.
- Deploy real infrastructure — Docker, VMs, SSH, tunnels, an LLM gateway — for ₹0
- Learn to orchestrate and judge an agent, not just chat with one
- Every action is tracked in a tamper-evident audit ledger, and mentors coach your daily log
Operating an agent is the skill
Chatting with AI is table stakes; operating an autonomous agent — provisioning it, governing it, judging its output — is the differentiator. Pie has you deploy your own personal AI agent on free cloud infrastructure, then run your daily miatz work through it. You orchestrate and judge; the agent executes.
Pie is an advanced module and is flag-gated: your admin enables it per tenant, so if you don't see it yet, it hasn't been switched on for your cohort.
The setup wizard
The guided wizard takes you through real infrastructure, for ₹0: Docker, VMs, SSH, tunnels, an LLM gateway, vector memory, the agent loop and governance. Each step teaches the concept first, then you do it on your own free cloud account. You'll also register the free LLM keys your agent will use, and finally register the pod itself — which gets a device token and a signing key.
Progress is tracked per step (skipped steps don't count as done), and you can return to the guide any time.
The dashboard
Once your pod is live, the Pie page becomes an operations console:
- Pod status — online/offline from heartbeats, provider, region, agent version
- Usage — tasks run through your Pie, first-pass rate, and estimated cost saved vs paid inference
- Audit ledger — a tamper-evident, hash-chained record of the pod's actions; entries signed with your pod's device key show as verified
The daily log and mentor loop
Each day, log what your Pie helped you ship — categorised as project, learning, testing or ops. This is where the coaching happens: mentors review your log and leave feedback on entries, so you improve not just what you build but how you direct the agent that builds it.
The journey, step by step
| Step | What you do | What you get |
|---|---|---|
| Start the setup wizard | Open Pie → Setup and work through the guided steps | Each step teaches the concept (Docker, VMs, SSH, tunnels, vector memory, governance) — then you do it for real |
| Add LLM keys | Register the free keys your agent will use from the setup page | Your pod has inference without paid APIs |
| Register your pod | Bring the agent online on your own free cloud | The dashboard shows status, heartbeat, provider, region and your device key |
| Run daily work through it | Log what your Pie helped you ship each day (project, learning, testing, ops) | A daily work log builds up, with usage stats: tasks, first-pass rate, cost saved |
| Close the mentor loop | Mentors review your log entries and leave feedback | You get coached on how you direct and judge your agent |
Frequently asked
I don't see Pie in my navigation — why?
Pie is an advanced, flag-gated module: an admin enables it per tenant. If the page isn't available, it hasn't been switched on for your cohort yet.
Does it really cost ₹0?
That's the design constraint: free cloud tiers for the pod, free LLM keys for inference. The setup wizard walks you through providers that fit. The dashboard even estimates the cost you've saved vs paid inference.
What is the audit ledger?
A tamper-evident, hash-chained log of your pod's actions. Your pod signs entries with its registered device key, and the dashboard shows which entries are cryptographically verified — governance is part of the lesson.
What does "online" mean on the pod card?
Your pod sends heartbeats; it shows as online when a heartbeat arrived in the last few minutes. Offline usually means the agent process or tunnel is down — redo the relevant setup step.
Related guides
A BYOK coach grounded in your own course, progress and reflections — powered by your own free key.
ReadHow to generate a free API key and connect it so Xook answers grounded in your learning.
ReadTrack assigned and self-set work with priority, impact and done-ness.
ReadFive live playgrounds for tokenization, chunking, embeddings, retrieval and sampling.
Read