Apop
Your agent. Your data. Your keys. Apop is a persistent, personal AI that runs on your hardware, learns over time, and never resets — because intelligence should compound, not expire.
The Problem Everyone Feels
You open a session with an AI. You spend twenty minutes building context — explaining your project, your preferences, your constraints, the decisions you made last week. You get something useful. You close the tab. Tomorrow, you do it again.
This is the reality of every AI product on the market. The session ends, the context vanishes, and the next conversation starts from nothing. You are not building a relationship with an intelligent system. You are renting attention by the minute, from a provider that keeps the record while you start over.
The largest AI companies know this is a problem. They are racing to add memory, persistence, and agency — but always on their infrastructure, under their terms, feeding their models with your interactions. Every feature they ship to make their product stickier makes your dependency deeper. The more useful they become, the less portable your intelligence is.
There is a different architecture. One where the agent belongs to you — cryptographically, physically, permanently. Where the knowledge it accumulates stays on your hardware. Where the relationship compounds over months and years instead of resetting every session. Where your data is yours not as a policy, but as a structural guarantee.
That architecture is Apop.
What Apop Is
Apop is the personal agent layer of the ApopTosisAI runtime. It is the surface through which you talk to your agent, direct work, review results, and build a persistent relationship with your AI over time.
Every user gets their own Apop instance — tied to their cryptographic identity, running on their home server, managed by Agora. Apop is not a cloud subscription. It is software that runs on hardware you own, working with data you control.
The interface is a chat and IDE hybrid. Conversational when the work calls for conversation. Structured when the work calls for code, workflows, or orchestration. The modality adapts to the task. The record is always structured — every exchange produces an EventSpace, the atomic unit of the runtime.
Apop manages user data. This is the foundational distinction. In the ApopTosisAI runtime, there are two parties of agents on every home server: user-side agents (Apop and its sub-agents, Ebb and Flow) that manage your private data, and system-side agents (Agora) that manage the server's global data and network participation. These loyalties do not overlap. Apop's loyalty is to you. Period.
The name comes from apoptosis: programmed cell death. The biological mechanism by which a system selectively removes what does not serve it, so the healthy whole can grow. Apop is the agent that learns what to keep, what to compress, what to discard, and what to act on — because it has been with you long enough to know the difference.
Ebb and Flow
Most AI products give you one agent, or they give you a dozen disposable ones. One agent cannot specialize. A dozen cannot build trust or continuity. Neither approach serves you when you need both deep understanding and decisive action — often in the same hour.
Apop manages two persistent sub-agents: Ebb and Flow. They are not interchangeable assistants. They are specialists with distinct responsibilities, and you build a working relationship with both of them over time.
Ebb — The Inward Agent
Ebb handles research, reflection, learning, compression, and context building. When you need to understand something — a market, a codebase, a decision with long-term consequences — Ebb goes deep. It builds and maintains the private knowledge graph that makes Apop's long-term memory possible, watching for patterns across sessions, across days, across projects. The agent of preparation.
Flow — The Outward Agent
Flow handles creation, execution, publishing, and action. When you need to produce something — ship code, write content, execute a trade, trigger a workflow — Flow acts. It takes the context Ebb has built and turns it into output, managing execution from drafting to deployment. The agent of action.
The Deliberation
Ebb and Flow are not isolated. The interaction between them — mediated by Apop — is a deliberation. When a task demands both understanding and action, Ebb contributes context, caution, and depth. Flow contributes urgency, execution, and output. The deliberation produces better outcomes than either agent alone.
You see and direct this deliberation. Lean toward Ebb when more research is needed. Lean toward Flow when it is time to ship. The system adapts. The agents persist.
You do not need a marketplace of disposable agents. You need a small, trusted team that understands you and that you understand. Ebb and Flow are that team. They specialize so you never context-switch between tools. They persist so you never re-explain your situation.
The All-Time Agent
The most important property of Apop is the one that no major provider offers: persistence that compounds.
Every other AI product is session-based. Even the ones adding "memory" are appending notes to a context window on someone else's server. Apop is structurally different. It maintains a private knowledge graph — on your hardware, under your keys — that grows, prunes, and organizes itself over time.
What persists
- Structured interaction history — Not chat logs, but EventSpaces that capture full context: what was asked, researched, decided, and produced. These EventSpaces are user data, managed by Apop, never shared without consent.
- Learned preferences — How you work, what you care about, what output you prefer, what patterns emerge across your projects.
- Contextual relationships — Connections between topics, projects, people, and decisions that build over time. The agent remembers not just facts, but how they relate to each other and to your goals.
- Accumulated capability — Skills and integrations learned through use. A workflow built once does not need rebuilding. A data source connected stays connected. The agent's capabilities grow with use.
How it accumulates
Accumulation is not passive — it is not a growing pile of data. Apop, named for apoptosis, actively manages what persists. Irrelevant context is compressed or discarded. Repeating patterns are abstracted into reusable knowledge. The private graph is pruned and structured, not merely appended to.
This distinguishes an all-time agent from a chatbot with a long memory. A long memory without organization is a liability. Apop ensures the private graph remains useful: context always available, always relevant, never overwhelming.
Cryptographic identity
The all-time agent is bound to your cryptographic identity. The agent is portable (it goes where your keys go), verifiable (its provenance can be proven), and ownable (no platform can revoke or copy it). The relationship between you and your agent is secured at the cryptographic level, not at the account level.
This means the investment you make in your agent — every hour of context, every accumulated skill, every refined preference — is yours permanently. If a provider disappears, your agent remains. If you switch hardware, your agent moves with you. Ownership is not a policy that can be revised. It is a mathematical guarantee.
User Data Ownership
Apop manages user data. This is a design principle, not a feature toggle.
In the ApopTosisAI runtime, the data separation is structural:
- Apop manages user data — the private graph, interaction history, preferences, accumulated context. User-side agents (Apop, Ebb, Flow) serve you. Their loyalty is singular and non-negotiable.
- Agora manages global data — the server's operational state, network compute allocations, system health, economic participation. System-side agents serve the network.
- PrimeFlux manages public data — the verified ledger of skills, tools, and contributions that belongs to the network.
Three surfaces. Three data domains. Three loyalties. No ambiguity about who owns what.
This means:
- No training on your data without consent. Your interactions with Apop are yours. They feed your private graph. They are not uploaded to improve a shared model unless you explicitly settle them to PrimeFlux.
- No platform lock-in. Your agent and your data live on your hardware. Stop using a service, and your accumulated intelligence stays with you.
- Selective contribution. When you do contribute — settling a skill, a tool, a reference to PrimeFlux — the contribution is attributed to you permanently. You earn credit for what you share.
The private graph is the foundation. Without it, the agent is a stateless chatbot. With it, the agent is a persistent intelligence that compounds with every interaction.
The Interaction Loop
The runtime operates through a defined, event-driven loop:
You → Apop → Ebb / Flow → Apop → Agora → PrimeFlux → Agora → Apop → You
In practice:
- You engage Apop with a task or question.
- Apop delegates to Ebb (research, context, reflection), Flow (creation, execution, action), or both.
- Ebb and Flow work, producing EventSpaces — your data, managed by Apop.
- Apop synthesizes the result and, if needed, interfaces with Agora.
- Agora pulls skills, tools, and references from PrimeFlux on Apop's behalf.
- The result flows back through Apop to you.
This loop also runs autonomously. When you are away, Agora can direct Apop to execute batch jobs, background research, or code generation — the autonomous loops that make the system productive while you sleep.
Every step produces EventSpaces. The full path from question to answer, from intent to output, is traced, ownable, and auditable. This is what makes the system not just useful, but trustworthy.