Symbiosis Core
In developmentFour contracts every Renvar product stands on.
Identity, Memory, Audit and Agent — the four contracts that allow a portfolio of products to behave as one intelligence rather than as a set of disconnected applications.
- Role
- Foundation for the whole portfolio
- Contracts
- Identity · Memory · Audit · Agent
- Build order
- First — before the products on top
- Design rule
- Products depend on the core, not each other
Overview
Most companies holding a portfolio end up holding a portfolio of integrations instead: every product re-implements authentication, every product invents its own notion of a user's history, and nothing composes. Symbiosis Core exists to prevent that outcome. It is built first, and everything else is built upon it.
The core exposes exactly four contracts. Core Identity issues capability tokens — scoped, expiring, revocable grants of authority that a product or an agent can present instead of a password. Core Memory is the sovereign, encrypted, category-isolated store of what a person has said and done, portable across models. Core Audit records every access to memory and every consequential decision in an append-only trail. Core Agent is the interface any reasoning process implements to participate — so a new agent is a new implementation, not a new platform.
The architectural rule is strict and deliberate: products depend on the core and never on one another. That is what makes it possible to add trust and verification to a messenger, or a care mode to a personal agent, without altering either product's internals.
Capabilities
Core Identity
Capability tokens rather than ambient permissions. Authority is scoped to an action, an amount and a window, and can be withdrawn without rotating anything else.
Core Memory
Encrypted personal memory with hard category isolation for health, finance and relationships, an off-the-record switch, and export that actually moves the context to another assistant.
Core Audit
Append-only record of who read what, which agent decided what, and on whose authority. Built for the moment a regulator or a customer asks.
Core Agent
One interface for reasoning processes. Multi-agent stacks — planner, executor, forecaster, critic — plug in without the core knowing what they do.