MellyCore AIOS Static product preview
MellyCore AIOS Local-first AI operating system

One command center. Every AI plane under operator control.

Explore models, agents, tools, context, and workspaces through one cinematic, provenance-bearing AI operations experience.

MellyCore keeps provider and runtime choices vendor-neutral while evidence, approvals, and safety boundaries stay explicit.

SystemLocal-first ControlOperator-controlled BuildStatic preview ActivityNo live providers or execution
Explore Command Center Explore AI Workspaces Static preview
Source Arena

Evidence before execution.

Representative snapshot
Context Committed sources Provenance retained
Routing policy Vendor-neutral Capability before cost
AIOS control MellyCore Operator approval required
Evidence Audit snapshot Source-linked records
Safety boundary Fail-closed No autonomous action
Data origin Simulated Static product projection
Command deck / visual readout STATIC PREVIEW
Context plane
REPOSITORY SNAPSHOT
Routing plane
FIXTURE DATA
Governance plane
APPROVAL-GATED

NO EXECUTION CAPABILITY

Representative Source Arena view of context, routing policy, operator approval, and evidence. It does not reflect live provider or runtime state.
Static system map

Command Center projection

A representative map of the product planes the Command Center can make legible. It is a static projection, not a connected runtime.

Context and routing

Committed context, explicit choice.

Source references and routing policy are shown as reviewable projections, not active system traffic.

  • ContextCommitted source records
  • RoutingCapability-aware policy
  • EvidenceSnapshot references
System calibration Static / source-backed

Measured operating state

Repository-backed disclosure for this representative Command Center projection.

Build classStatic product preview
Provider stateNot connected
ExecutionNo live activity
Workspaces10 planned
RoutingCapability before cost

Provenance Committed project context and static preview baseline.

Control boundary

Approval remains human.

Agents, runtime concepts, tools, and observability are legible without implying autonomous execution.

  • RuntimeFramework-neutral view
  • ToolsReference-only surfaces
  • ApprovalRequired before action
  1. Context
  2. Routing
  3. Runtime
  4. Tools
  5. Approval
  6. Execution
  7. Evidence

Static product projection. This flow describes how the product organizes concepts; it does not indicate live execution, routing, or telemetry.

Static conceptual projection

Runtime Constellation

A framework-neutral way to understand potential runtime boundaries around MellyCore—without asserting installation, support, connection, execution, or authorization.

Projection 01 / 06 Framework bridge MellyCore Projection, not runtime owner
Claude Code StateReference
OpenAI Agents SDK StateReference
LangGraph StateReference
CrewAI StateReference
AutoGen StateReference
MellyCore Custom StatePlanned projection
Framework-neutral center MellyCore projection Not a runtime owner
  1. Claude CodeReference
  2. OpenAI Agents SDKReference
  3. LangGraphReference
  4. CrewAIReference
  5. AutoGenReference
  6. MellyCore CustomPlanned projection
Reference — canonical framework term, not implementation evidence Planned projection — product direction only

Displayed does not mean installed, supported, connected, running, or authorized. The constellation is a static framework-boundary projection with no provider calls or runtime activity.

Runtime / framework projection Operator capability field AI Workspaces
Layer 02 / Capability destinations 10 planned workspaces

AI Workspaces

Ten operator-oriented environments for research, building, communication, media, and local AI direction—presented as one planned product ecosystem, not active applications.

Static ecosystem preview. These planned surfaces have no workspace backends, provider access, account connections, downloads, or execution.
Wave 1 / 04 workspaces

Primary capability field

Flagship technical work plus evidence synthesis, comparison, and bounded collaboration.

W01 / Flagship stage

Coding / Runtime Studio

A planned operator workspace for shaping code and runtime tasks with context, approvals, and evidence kept in view.

Planned workspace
Context frame
Task
Representative brief
Runtime
Not selected
Evidence
Source references
Static work surface Bounded implementation outline
  1. 01Define the requested boundary
  2. 02Review committed context
  3. 03Prepare evidence for operator review

No editor, terminal, or execution control is present.

Authority boundary Operator review

Consequential action remains outside this static composition.

No execution
Workspace
Technical capability destination
State
Planned / static preview
Authority
Operator-controlled
Origin
Committed product roadmap
W02 / Evidence synthesis

Deep Research

A planned workspace for turning a bounded question and visible sources into a traceable report.

  1. Plan
  2. Read
  3. Synthesize
  4. Report
State
Planned
Output
Source-linked report
Origin
Roadmap direction
W03 / Parallel evaluation

Compare Arena

A planned comparison surface for reviewing alternatives side by side before a human decision.

View AResponse placeholder Blind review View BResponse placeholder
State
Planned
Decision
Human evaluation
Origin
Roadmap direction
W04 / Bounded roles

Multi-Agent Crew

A planned workspace for composing explicit roles without turning framework concepts into autonomous authority.

SupervisorPlannerResearcherReviewer
State
Planned
Mode
Bounded collaboration
Runtime
Reference only
Wave 2 / 03 workspaces

Commercial capability plates

Concrete communication and media outcomes, with every consequential boundary stated at point of use.

W05 / B2B communication

Email AI

Planned triage, summaries, action extraction, and reply drafting for reviewable business communication.

State
Planned / no account connection
W06 / Media understanding

Video Intelligence

A planned surface for scene, topic, summary, and highlight concepts—without uploaded media or inference.

ScenesTopicsSummaries
State
Planned / no media connected
W07 / Operator interface

Voice

A planned interface path for spoken intent. It does not own authority and no microphone is connected.

  1. Voice
  2. STT
  3. Intent
  4. Control Plane
  5. Result
  6. TTS
State
Planned interface / no input
Wave 3 / 03 workspaces

Local + multimodal expansion

One grouped future capability field for visual creation and operator-controlled local-model direction.

W08 / Multimodal

Image Studio

Planned image orchestration, comparison, lineage, and controlled editing concepts. No generation occurs here.

W09 / Local artifacts

Model Downloader

Planned model discovery, provenance, licensing, storage, and hardware-fit review. No downloads are available.

W10 / Local runtime

Ollama Manager

Planned local-runtime visibility. Ollama, local models, and detected hardware are not assumed present.

Group state
Planned expansion
Authority
Operator-controlled
Local runtime
Not detected or asserted
Origin
Committed product roadmap

Presentation hierarchy, not activation order. Visual presence does not establish implementation, provider support, workspace access, synchronization, downloads, or execution.

Static file list — no runtime access

Shared Context / Memory Layer

A committed-context projection for the static preview. Metadata labels are shown; this page does not read from a runtime memory system.

Source directory
shared_context/
Projected records
9 of the committed set
  • AGENT_HANDOFF.md

    A committed handoff record, shown as metadata only.

    Handoff
  • RUN_QUEUE.md

    A committed planning record, not a live task feed.

    Queue
  • SAFETY_CONTRACT.md

    Committed safety constraints and operator boundaries.

    Contract
  • ROADMAP.md

    A committed roadmap reference, shown statically.

    Roadmap
  • MODEL_ROUTING.md

    A vendor-neutral routing reference, shown statically.

    Routing
  • DESIGN_SYSTEM.md

    Visual direction and design constraints.

    Design
  • PROJECT_STATE.md

    A committed project-state reference, shown statically.

    State
  • VALIDATION.md

    Baseline validation commands and checks.

    Validation
  • TOOLING.md

    Tool and workspace definitions.

    Tooling
Knowledge & operations graph Derived product view No live ingestion

Knowledge & Operations Graph

A derived, representative view of how context, decisions, systems, runtime concepts, and evidence become understandable together. It does not own canonical truth.

FIXTURE DATA REPOSITORY SNAPSHOT Derived fixture Representative clusters No live ingestion
Central relationship view

Representative relationship map

Five provenance-bearing fields and five selective relationships derived from committed project records—not live ingestion.

SIMULATED SURFACE NO RUNTIME FEED Derived static view. This graph projects source records; it is not a source of truth.
Field 01 Product foundation MellyCore product vision Product vision Design constraints OriginPROJECT_STATE
Field 02 / canonical source Canonical context Committed records Decisions Source references Originshared_context/
Field 03 / derived plane Operations projection Derived graph Runtime concepts Tools OriginRepresentative view
Field 04 / governed crossing Approval boundary Operator approval Fail-closed policy No autonomous action OriginSAFETY_CONTRACT
Field 05 / retained proof Evidence records Audit snapshot Review trail Validation record OriginRepository records
Center of gravity Canonical context Committed records remain the source; this graph only projects their relationships.
  1. 01
    Product foundation projects from canonical context

    Origin — PROJECT_STATE

  2. 02
    Canonical context references committed source records

    Origin — shared_context/

  3. 03
    Context projects a derived operations view

    Origin — representative static view

  4. 04
    Operations crosses an operator approval boundary

    Origin — SAFETY_CONTRACT

  5. 05
    Evidence records retain the review trail

    Origin — repository records

Textual alternative: product foundation projects from canonical context; committed context references source records and projects a derived operations view; operations cross an operator approval boundary; evidence records retain the review trail. Runtime and tool concepts remain projections. The graph owns none of these records.

Relationship legend

  • projects from
  • references
  • approval boundary
  • evidence trail

Projection safeguards

  • no-live-ingestion
  • no-runtime-owner
  • no-provider-call
  • no-autonomous-action

This view makes projection boundaries visible. It does not validate, connect, or execute against a runtime system.

Data origin and evidence

  • Committed project context
  • Homepage product specification
  • Static review snapshot

Data origin: simulated/representative projection of committed records. No live graph generation, runtime ingestion, backend/API/database, external integration, or execution exists here.

Supervised workflow

Observe → analyze → recommend → approve

Control flow shown as a static design contract. Implementation and execution are outside this preview.

  1. 01ObserveReviewable context
  2. 02AnalyzeStructured reasoning
  3. 03RecommendHuman-readable proposal
  4. 04ApproveOperator decision gate
Static roadmap — updated manually

Roadmap Orbit Map

Project phases as orbital progression. Inner orbits are complete; outer orbits come later. Honest state — nothing implied beyond it.

Bootstrap Complete

Repo scaffold, shared context files, safety contract.

Cross-agent smoke Pending

Verify shared context across all agents and tools. Not yet run.

Design & spec docs Complete

Design system, homepage spec, frontend scaffold plan.

Static website Current static slice

The current static showcase slice; it does not imply a live product surface.

MVP demo Planned

Static prototype with richer navigation and visual polish.

Runtime / provider integrations Later — approval gated

Requires explicit approval before activation. Not started.

  1. Bootstrap Complete

    Repo scaffold, shared context files, safety contract.

  2. Cross-agent smoke Pending

    Verify shared context across all agents and tools. Not yet run.

  3. Design & spec docs Complete

    Design system, homepage spec, frontend scaffold plan.

  4. Static website Current static slice

    The current static showcase slice; it does not imply a live product surface.

  5. MVP demo Planned

    Static prototype with richer navigation and visual polish.

  6. Runtime / provider integrations Later — approval gated

    Requires explicit approval before activation. Not started.

Safety contract reference

Safety / Control Layer

Explicit safety constraints, visible at every layer. Safety is architectural, not decorative.

No secrets No keys No unsupervised runs No copied workspace
  • No secrets in the repository.
  • No real API keys committed.
  • No provider tokens in tracked files.
  • No environment secret values in version control.
  • No destructive git operations without explicit approval.
  • No publishing or hosting without explicit approval.
  • No MellyTrade mutation from MellyCore AIOS.
  • No wholesale import of the GLM workspace — reference only.
  • No .git import from reference workspaces.
  • No database files committed.
  • No local runtime state committed.
  • Static-first before any runtime integration.
  • Honest validation — no fake status indicators.

The GLM/Z.ai workspace is reference only. No GLM files have been copied into this repository. Visual inspiration is acknowledged; wholesale import is forbidden.

MellyCore AIOS is standalone and separate from MellyTrade. This project contains no trading runtime, no broker connectivity, and no execution surfaces of any kind.

Controlled surfaces — no autonomous runs

Tooling Layer

Representative workflow references for the product architecture. No tool is launched, connected, or operated from this page.

Editor, terminal, and source surfaces

  • Warp Safe terminal workflows, operator prompts, repeatable runbooks.
  • Zed Editor-agent workspace, fast code navigation, local editing ergonomics.
  • VS Code Broadly compatible editor workspace with extension-compatible docs.
  • GitHub Remote source control, issues, PRs, review trail, collaboration history.

Agent surfaces

  • Claude Code Architecture, reasoning, documentation, review workflows.
  • Codex Implementation, validation, local git hygiene, PR preparation.
  • GLM 5.2 Drafting, iteration, exploration, reference implementation ideas.
  • ChatGPT Strategy, context synthesis, prompt generation, planning, memory.
Surface state
Reference only — not launched or connected
Operator Channel

Move from product proof to implementation scope.

MellyCore demonstrates how governed AI systems can be shaped across research, coding, communication, media, and local-model workflows. This static build has no reviewed inquiry or onboarding endpoint.

Contact limitation: direct project inquiry is not configured here; no message, account, or access request can be submitted.

Channel state
Local reference only
System effect
None
Contact endpoint
Not configured
Review product specification Open static dashboard preview
These links open existing local project surfaces. They do not trigger execution or external communication.