Agora
Your home server is not an island — it is a node. Agora turns personal compute into a participant in a shared network, orchestrating your devices, managing your agents, and earning through contribution.
The Shift That Is Already Happening
People bought their own phones. Then their own laptops. Then their own tablets. Now they are buying their own servers.
The hardware to run AI locally — Apple Silicon, consumer GPUs, edge devices — is already in people's homes. The models are getting smaller and more capable every month. The tools to serve them — Ollama, MLX, Docker — are mature, open-source, and improving fast. The infrastructure for personal AI compute is not a future promise. It is a present reality.
What is missing is the software that turns that hardware into something more than a hobbyist setup.
Right now, if you run models at home, you get inference on one machine. You do not get orchestration across your devices. You do not get participation in a shared compute network. You do not get economic return on the compute you are already running. You get isolation where you could have leverage.
Meanwhile, the largest AI providers are racing in the opposite direction — centralized compute, centralized data, centralized control — shipping features that make their platforms stickier while making your dependency deeper. Every new capability they offer comes with the same structural condition: your data on their servers, your compute on their terms, your intelligence locked inside their platform.
Agora is the other path. The one where your hardware works for you — and works for a network that works for you in return.
What Agora Is
Agora is one of three surfaces in the ApopTosisAI runtime. Where Apop manages your private data and personal agent, and PrimeFlux manages the public ledger of verified skills and contributions, Agora manages the global data of the home server — its operational state, its network participation, its compute economics, and its relationship to the rest of the network.
Named for the ancient Greek marketplace — the public square where exchange happens — Agora is a unified concept with three dimensions:
Agora as agent. The highest-level intelligence on your home server. It reasons about resource allocation, job scheduling, network participation, and compute economics. Where Apop thinks about what you want, Agora thinks about what the system needs. It is the counterpart: one agent serves the person, the other serves the network, and both coexist on the same hardware.
Agora as interface. Mission control for your home compute. The GUI where you see what your node is doing, what models are running, how your devices are connected, what you are contributing to the network, and what you are earning.
Agora as principle. The marketplace itself — the mechanism by which home servers become more than isolated machines. Fair exchange, contribution and reward, compute as a shared resource. Agora is the space where private compute meets public participation.
These are not separate products. They are facets of the same thing. You interact with Agora through mission control. Behind the interface, the Agora agent orchestrates. The orchestration follows the Agora principle: contribute, participate, earn.
Two Parties of Agents
The core architectural insight of the ApopTosisAI runtime is that every home server runs two parties of agents with different loyalties and different data domains.
User-side agents: Apop, Ebb, and Flow. They serve you. They manage your data — the private graph, interaction history, preferences, accumulated context. Their loyalty is to the person.
System-side agents: Agora. It serves the network. It manages global data — the server's operational state, network compute allocations, economic participation, system health. Its loyalty is to the network.
This is not a conflict. It is a design principle that solves one of the hardest problems in personal AI: how do you build a system that benefits the individual and the network simultaneously, without either compromising the other?
The answer is structural separation. Two parties, two data domains, two sets of responsibilities, coexisting on the same hardware. Apop never exposes your data to the network without consent. Agora never prioritizes the network at the expense of your explicit directives. The boundary is architectural, not policy-based.
The interaction between Apop and Agora is the fundamental loop of the system. When Apop needs compute, Agora allocates it. When Apop needs a skill from PrimeFlux, Agora retrieves it. When your interaction produces an EventSpace worth settling to the ledger, Agora manages the settlement. When the network needs compute from your node, Agora decides how much to contribute based on current load and your directives.
Apop is your personal assistant. Agora is your building's superintendent. Different masters, same building, working together.
Home Compute and Device Orchestration
A home server is not one device. It is a cluster.
A Mac Studio in the office. A MacBook in the living room. An iPhone in your pocket. An Apple TV on the screen. Each device has compute capacity. Each runs different workloads. Together, they are more capable than any one of them alone.
Agora orchestrates across this cluster. It knows what hardware is available, what models are loaded, what inference capacity exists, and how to route work to the right device. Heavy reasoning goes to the most powerful machine. Always-on monitoring goes to low-power devices. You do not manage this manually. Agora does.
Integration points
Agora builds on existing, proven infrastructure — not aspirational dependencies, but shipping software that works today:
- Ollama — Local model serving. Agora manages which models are loaded, on which devices, and routes inference to the right instance.
- MLX — Apple Silicon machine learning framework. For models optimized for Apple hardware, MLX provides the inference layer that Agora orchestrates.
- Docker — Containerization. Each surface of the runtime — Apop, Agora, PrimeFlux — runs in its own container with its own dependencies, managed by Agora.
- Tailscale — Private mesh networking. Encrypted, zero-config connectivity across all devices in the cluster — and eventually, between nodes in the wider network.
The development strategy is integration, not reinvention. These tools solve their respective problems well. Agora's job is to unify them into a single, coherent experience.
Nodes Sharing Compute
A single home server is useful. A network of home servers is transformative.
When multiple nodes run Agora, they can share compute. A task exceeding one node's capacity distributes across the network. A model loaded on one node serves inference for another. Idle compute on one machine contributes to a job on another.
This is the network dimension of Agora — the marketplace principle applied to compute itself. Nodes contribute what they have and draw what they need. The protocol tracks contributions and consumption, and the economics flow through PrimeFlux as verifiable records.
Net zero memory
One of Agora's most ambitious goals: the ability for the network to function as if it has access to the combined memory and compute of all participating nodes. No single node holds everything. The network, taken as a whole, does.
This is not new in distributed computing. It is new in the context of personal AI. Today, running a large model requires a machine powerful enough to hold it — or a cloud subscription. In the Agora network, the combined memory of participating home servers runs models and workloads that no single node could handle alone.
The mechanisms — node discovery, work partitioning, result aggregation, trust between nodes — build on established patterns in distributed computing, adapted for the specific constraints of personal AI hardware. The principle is fixed: your home server is not an island. It is a node in a collective that is more capable than the sum of its parts.
The network-effects dynamic is structural: every node that joins makes every other node more capable. The value of participation increases with scale — and the switching cost increases with accumulated contribution.
The Heartbeat
Agora keeps the system alive through a heartbeat — a continuous, event-driven cycle that manages jobs, monitors health, and maintains the connections between all parts of the system.
The heartbeat makes the system always-on:
- Device health — which devices are online, their load, available models, inference capacity.
- Autonomous jobs — tasks that run without user direction: batch processing, background research, code generation, model updates, data synchronization.
- Interaction with Apop — when you are away, Agora directs Apop to execute scheduled work, triggers Ebb for background research, queues tasks for Flow when resources free up.
- Network participation — pinging the broader network, updating node availability, processing incoming compute requests.
The distinction between user-directed and autonomous work is critical. When you are present, Apop drives and Agora supports. When you are away, Agora drives and Apop executes. The heartbeat governs this transition — not a literal timer, but an event-driven cycle responsive to the state of the system.
This is where autonomous agent frameworks — open-source orchestration tools and the autonomous capabilities now shipping from major AI providers — integrate with the runtime. Agora orchestrates these loops: when to start them, when to stop them, what resources they receive, how their output is recorded as EventSpaces.
Your server works while you sleep. It researches, it processes, it learns, it earns. This is the mechanism by which a home server generates passive income — not through speculation, but through verified, continuous contribution to a network that compensates real work.
Mission Control
The Agora interface is mission control for your home compute — the GUI where the operational reality of your node becomes visible and manageable.
What mission control shows:
- Your devices — what is online, running, loaded, and how hard each device is working.
- Your agents — Apop, Ebb, Flow, and any autonomous loops. Their status, current tasks, resource consumption.
- Your network — your node's connection to the broader network. What you contribute, what you consume, your peers.
- Your economics — what your node has earned through compute contribution and skill settlement. The tangible result of participation.
- Your jobs — active and queued work. Batch tasks, background research, scheduled maintenance.
Mission control is not a monitoring dashboard. It is a control surface. Direct Agora to prioritize certain jobs. Limit what your node contributes to the network. Load or unload models. Set directives for autonomous operation.
The design philosophy: managing a home AI server should feel like managing a home. Set preferences. The system handles the details. Check in when you want to. Form is function — the interface is as intuitive as the home screen on your phone, not a systems administration console.
Agora as Ecosystem
There is a reason Agora is named for the marketplace.
The practical surface — mission control, orchestration, job management — is what you interact with daily. But the deeper role is structural: Agora is the interface between private and public, between compute that sits idle and compute that creates value, between a home server and a network.
Consider the analogy: X has Grok. An AI that lives inside the platform, trained on the platform's data, serving the platform's users. Grok is the system-level agent for X.
Agora is that concept applied to the home server. It is the system-level agent for your node. It understands your compute, participates in the network, grows as the network grows. But unlike Grok, Agora does not belong to a platform. It runs on your hardware, serves your interests alongside the network's, and operates under a protocol rather than a corporate policy.
This is the fundamental positioning: Agora is what happens when the "big data" model of AI — centralized compute, centralized data, centralized control — meets the reality that people want to own their infrastructure. The models are small enough to run at home. The hardware is powerful enough to serve them. The only missing piece is the software that connects those home servers into something larger.
That software is Agora. And the market for it grows with every person who decides they want their own compute — which is a trend that has been accelerating since the first personal computer, and is now reaching its logical conclusion.