Marinext documentation
Marinext is not another dashboard wrapped around chat. It is a personal operating system for attention, memory, automation, and execution: it watches the right inputs, reasons over them, acts through agents and workflows, and turns what happened into reusable memory.
Heartbeat, Today, schedules, and integrations keep the system aware of new work.
Agents, workflows, approvals, runs, and codebases move from decision to execution.
Memory, graph relationships, errors, lessons, and reflection keep improving the system.
Why it matters
The app is designed around an execution loop: capture signals, evaluate context, decide what matters, act with guardrails, and preserve the lesson. That makes it closer to a living operations layer than a static productivity tool.
Structure
This first version creates the spine. Each app menu has a dedicated section below, and each section can later become a full documentation article with workflows, screenshots, examples, and troubleshooting.
The conversational command center for asking, planning, drafting, and executing through the Second Brain.
Chat keeps conversation context tied to sessions, can surface workflow activity, and is the natural place to ask the system to reason across memory and operations.
A daily operating view for inbox, calendar, tasks, content deadlines, heartbeat status, and active work.
Today answers the practical question: what needs attention now, and what changed since the last heartbeat?
The evaluation cockpit for whether the Second Brain can notice, decide, act, and learn safely.
Brain includes Program OS gates, evidence, memory quality signals, pending actions, and governance checks.
The roster of specialist operators that handle domains, tasks, tools, and recurring responsibilities.
Agent pages show role, ownership, activity, memory links, and how each agent participates in workflows.
The automation builder for repeatable multi-step work across integrations, agents, approvals, and memory.
Workflows turn repeatable processes into inspectable runs with versioned definitions and operational history.
Cost and reliability intelligence for models, runs, workflows, agents, and auto-fix performance.
Analytics helps the system stay useful without becoming expensive or opaque.
A relationship map across memory, runs, agents, workflows, codebases, and useful evidence.
Graph is where scattered knowledge becomes navigable context instead of isolated notes.
The calendar for recurring automations and planned system work.
Schedules make proactive behavior visible: what runs, when it runs, and what it is allowed to do.
The safety queue for actions that need a human decision before the system proceeds.
Approvals keep high-impact actions explicit while preserving the flow of automated work.
The registry of repositories, worktrees, spend, and coding surfaces connected to the Second Brain.
Codebases make software work traceable across runs, fixes, branches, and deployed projects.
The execution ledger for workflows, heartbeats, agent jobs, and recovery attempts.
Runs show what happened, where it happened, what it cost, and what evidence was produced.
The focused inspection view for the active or selected run.
Run Detail is useful when something is executing now and the operator needs timeline, state, artifacts, and recovery controls in one place.
The reliability desk for failures, clusters, lessons, and auto-fix candidates.
Errors turns incidents into reusable learning instead of one-off debugging.
The control panel for account, integrations, chat defaults, agent settings, budgets, and advanced operations.
Settings is where the app exposes configuration without mixing it into everyday work surfaces.
Search for a command to run...