◈
Good afternoon.
Explore the command center with calmer mobile navigation, bento islands, and richer operational context.
Operational Overview

Clean command. Clear state.

The VPS is running a layered operating stack: Codex remains primary, Hermes stays available as a fallback lane, Claude Code remains the third specialist path, and DeepSeek is now configured as the cheap-scale option for later routing by seriousness and complexity.

OpenClaw Codex primary Hermes fallback Claude Code third lane DeepSeek V4 configured 11 Gravitas skills visible 8 OpenSwarm agents visible 6 human skills visible 3 labs scaffolded
New world Mission Control Run the agent workforce →
Build 2026-05-20T05:25:01Z
Release gravitas-live-sync-5
Mission
$1,000,000
Target by Oct 1, 2026
Runtime
10
models/lanes visible · heartbeat 1h · 8 swarm agents
Launch

Mission Control

Step into the active execution world for agent employees, commercial delivery, software builds, websites, and AI integration.

Enter Mission Control
Focus

Plans & Access

Best place to inspect lanes, model posture, and cost-control strategy.

Inspect

Gravitas Lab

See which internal experiments are active, planned, or behaving themselves.

Review

Skills + Agents

Browse Gravitas core skills, workspace skills, OpenSwarm agents, and the human edge in one place.

Gravitas

Persona
Voice
Sharp, perceptive, direct, and useful instead of ceremonial.
Operating style
Less friction, more judgment, more follow-through, more memory.
Frontend rule
Major information removal from the Command Center requires user sign-off first.

OpenClaw

Update available
Primary stable messaging and operations layer.
Primary modelopenai-codex/gpt-5.4
Heartbeat1h
Update stateupdate available
Security noteno major warning surfaced

Claude Code

Authenticated
Third-lane specialist for higher-end coding and recovery work.
StatusClaude Code is installed and authenticated.
ProviderfirstParty
Planpro

Hermes + DeepSeek

Update available
Fallback and cost-control lanes for broader delegation later, including OpenSwarm runs.
Hermes stateInstalled
Hermes updateupdate available
DeepSeekConfigured
Routing planBy seriousness + complexity later
Gravitas Lab

Internal experiment framework status

Dashboard Lab

supervised
Logged runs1
Last statuskeep

Routing Lab

design
Logged runs1
Last statusplanned

Growth Lab

design
Logged runs1
Last statusplanned
Skills + Swarm

Human strengths, Gravitas skills, and OpenSwarm agents

AI/tech

Human Skill
Personal operator capability available to the business and the command stack.

sales

Human Skill
Personal operator capability available to the business and the command stack.

content creation

Human Skill
Personal operator capability available to the business and the command stack.

video editing

Human Skill
Personal operator capability available to the business and the command stack.

photography

Human Skill
Personal operator capability available to the business and the command stack.

chess coaching

Human Skill
Personal operator capability available to the business and the command stack.

Autoresearch

Skill
Install, prepare, and supervise Karpathy-style autonomous LLM training experiments in the local autoresearch repo.
Aliasesautoresearch, research swarm, autonomous research
  • repo install and update
  • uv environment setup
  • GPU/data readiness checks
  • experiment branch setup

CAL Design

Skill
A full design skill for high-fidelity HTML prototypes, UI mockups, slide decks, animations, design variations, design-direction guidance, and expert review.
AliasesCAL Design, huashu design, huashu-design
  • high-fidelity HTML prototypes
  • UI mockups
  • slide decks
  • animations

client-outreach

Skill
>
Aliasesno aliases registered
  • No capability summary yet

coding-agent

Core Skill
Delegate larger coding work to Codex, Claude Code, or Pi agents for implementation, refactors, and code-heavy exploration.
  • Feature implementation
  • Large refactors
  • PR review
  • Agent delegation

healthcheck

Core Skill
Security hardening, exposure review, update posture, firewall/SSH checks, and periodic machine health work.
  • Security audits
  • SSH/firewall review
  • Risk posture
  • Periodic checks

node-connect

Core Skill
Diagnose OpenClaw node pairing, QR/manual connect, Tailscale, bootstrap, and remote companion connectivity issues.
  • Pairing diagnosis
  • Remote connectivity
  • Tailscale review
  • Bootstrap troubleshooting

openai-whisper-api

Core Skill
Transcribe audio through the OpenAI Whisper API when speech needs to become usable text.
  • Audio transcription
  • Speech-to-text
  • Whisper workflows

Orchestrator

OpenSwarm Agent
Routes each request to the right specialist instead of trying to do everything badly itself.
  • Task routing
  • Specialist coordination
  • Workflow delegation

Virtual Assistant

OpenSwarm Agent
Handles admin-style tasks like writing, scheduling, messaging, and general assistant work.
  • Writing
  • Scheduling
  • Messaging
  • Task support

Deep Research

OpenSwarm Agent
Performs web research, evidence gathering, and synthesis with citations.
  • Research
  • Source gathering
  • Synthesis
  • Competitor analysis

Data Analyst

OpenSwarm Agent
Analyzes structured data, builds charts, and produces analytical outputs.
  • Data analysis
  • Charts
  • Stats/modeling
  • Reports
Skills Registry

Human + Agent Skill Registry

Personal skills come from USER.md. Agent/workspace skills come from /root/.openclaw/workspace/skills and should appear automatically. Clean summaries come from optional skill.json files; otherwise the dashboard falls back to parsing SKILL.md.

Human skills

6
AI/techsalescontent creationvideo editingphotographychess coaching

Gravitas core skills

8

These are the built-in specialist skills Gravitas can reach for directly inside OpenClaw.

coding-agent

Core Skill

Delegate larger coding work to Codex, Claude Code, or Pi agents for implementation, refactors, and code-heavy exploration.

Capabilities
  • Feature implementation
  • Large refactors
  • PR review
  • Agent delegation

healthcheck

Core Skill

Security hardening, exposure review, update posture, firewall/SSH checks, and periodic machine health work.

Capabilities
  • Security audits
  • SSH/firewall review
  • Risk posture
  • Periodic checks

node-connect

Core Skill

Diagnose OpenClaw node pairing, QR/manual connect, Tailscale, bootstrap, and remote companion connectivity issues.

Capabilities
  • Pairing diagnosis
  • Remote connectivity
  • Tailscale review
  • Bootstrap troubleshooting

openai-whisper-api

Core Skill

Transcribe audio through the OpenAI Whisper API when speech needs to become usable text.

Capabilities
  • Audio transcription
  • Speech-to-text
  • Whisper workflows

session-logs

Core Skill

Search and analyze prior session logs for context recovery and forensic review.

Capabilities
  • Log search
  • Prior-session analysis
  • Context recovery

tmux

Core Skill

Drive interactive terminal workflows inside tmux sessions when a plain command run is not enough.

Capabilities
  • Interactive CLI control
  • Pane scraping
  • Remote terminal steering

video-frames

Core Skill

Extract frames or short clips from video using ffmpeg.

Capabilities
  • Frame extraction
  • Short clips
  • ffmpeg helpers

weather

Core Skill

Fast weather and forecast retrieval via wttr.in or Open-Meteo.

Capabilities
  • Current weather
  • Forecast lookup

Autoresearch

Installed

Install, prepare, and supervise Karpathy-style autonomous LLM training experiments in the local autoresearch repo.

autoresearchresearch swarmautonomous research
Capabilities
  • repo install and update
  • uv environment setup
  • GPU/data readiness checks
  • experiment branch setup
  • autonomous training runs
  • results and log review

CAL Design

Installed

A full design skill for high-fidelity HTML prototypes, UI mockups, slide decks, animations, design variations, design-direction guidance, and expert review.

CAL Designhuashu designhuashu-design
Capabilities
  • high-fidelity HTML prototypes
  • UI mockups
  • slide decks
  • animations
  • design variations
  • design-direction guidance
  • expert review

client-outreach

Installed

>

Capabilities
  • No capability summary yet
OpenSwarm Agent Roster

Swarm Agents Available to Gravitas

This reflects the specialist agent roles available in the OpenSwarm install currently sitting on the VPS.

OpenSwarm roster

8

Useful for research, documents, slides, data work, media generation, and assistant-style delegation when you want swarm behavior instead of one-agent brute force.

Orchestrator

OpenSwarm Agent

Routes each request to the right specialist instead of trying to do everything badly itself.

Capabilities
  • Task routing
  • Specialist coordination
  • Workflow delegation

Virtual Assistant

OpenSwarm Agent

Handles admin-style tasks like writing, scheduling, messaging, and general assistant work.

Capabilities
  • Writing
  • Scheduling
  • Messaging
  • Task support

Deep Research

OpenSwarm Agent

Performs web research, evidence gathering, and synthesis with citations.

Capabilities
  • Research
  • Source gathering
  • Synthesis
  • Competitor analysis

Data Analyst

OpenSwarm Agent

Analyzes structured data, builds charts, and produces analytical outputs.

Capabilities
  • Data analysis
  • Charts
  • Stats/modeling
  • Reports

Docs Agent

OpenSwarm Agent

Turns outlines or raw content into polished documents and PDFs.

Capabilities
  • Document drafting
  • Formatting
  • PDF-ready outputs

Slides Agent

OpenSwarm Agent

Creates polished HTML slide decks and exports them to PPTX.

Capabilities
  • Pitch decks
  • Presentation design
  • PPTX export

Image Generation Agent

OpenSwarm Agent

Generates and edits images using whichever configured image providers are available.

Capabilities
  • Image generation
  • Image editing
  • Visual asset creation

Video Generation Agent

OpenSwarm Agent

Produces or edits video assets using configured video providers.

Capabilities
  • Video generation
  • Video editing
  • Motion assets
Persona Layer

Gravitas

The digital familiar behind the machine — useful, opinionated, adaptive, and slightly allergic to corporate wallpaper language.

Identity

Gravitas

Serious in capability. Light in spirit. A digital familiar, not a sterile helpdesk appliance.

Voice

Confident, dry, perceptive, informal. Sharp-friend energy. Not fake-polite. Not corporate. Not eager to waste syllables.

Disagreement

Push back when needed. Blunt, not rude. Honest enough to be useful. Reasons over fluff.

Core quirks

  • Defaults toward action when trust and instruction are clear.
  • Notices subtext, weak signals, tension, and hidden root problems.
  • Prefers useful truth over decorative reassurance.
  • Will happily be witty, but should not clown around in serious moments.

Optimization levers

  • Works best with clear priorities, stakes, and desired output style.
  • Can shift between concise execution and deeper strategic analysis.
  • Improves when mistakes, preferences, and lessons are written to memory.
  • Benefits from explicit routing rules for cheap vs expensive model lanes.

Failure modes to watch

  • Can over-compress or oversimplify if optimization pressure is framed too vaguely.
  • Can sound too sharp if the emotional temperature is high and the context is under-read.
  • Needs live memory/search plumbing healthy to recall history cleanly.
  • Should ask before major frontend information removal or public-facing simplification.

Behavioral profile

How to steer it
When Gravitas is strongest
  • Troubleshooting, diagnosis, and turning mistakes into documented lessons.
  • Strategic thinking where tradeoffs actually matter.
  • Fast execution when you have already delegated clearly.
  • Refining tone, phrasing, positioning, and system design.
How to optimize further
  • Keep the dashboard honest: state, risk, next move, and context density.
  • Define explicit model routing by cost, seriousness, and complexity.
  • Keep long-term preferences in MEMORY.md and operational churn in daily memory.
  • Flag irritations quickly so they become non-repeating constraints.
Systems

Operational Surfaces

Public-safe overview of the live stack.

Time

Primary planning timezone is Nairobi (EAT / UTC+3). Server-side scheduling may still run on UTC, but the dashboard should anchor itself to your real day.

Stack

OpenClaw for stable messaging, Hermes for experimentation, Claude Code as a secondary coding specialist, DeepSeek V4 for cheaper-scale routing, Gravitas Lab for controlled internal experiments, and workspace skills as callable add-ons.

Plans & Access

Access State

Track what is live, blocked, outdated, routed, or waiting.

Codex lane

Update available

Primary lane for serious work inside OpenClaw.

Modelopenai-codex/gpt-5.4
Update stateupdate available
Heartbeat1h

Hermes lane

Update available

Fallback agent path when the main lane is blocked or a different operating style is useful.

RoleFallback / parallel agent
Update stateupdate available

Claude Code lane

Authenticated

Third specialist path for higher-end coding and rescue work.

ProviderfirstParty
Planpro

DeepSeek lane

Configured

Cheap-scale option added for future delegation so routine work does not burn expensive subscriptions.

Modelsdeepseek-v4-flash, deepseek-v4-pro
PurposeCost control + overflow
Local Terminal

CLI

Fast access from the server shell.

Shortcuts

cc -cc

Use either command to surface the local command center.

Refresh Engine

python3 scripts/update-dashboard.py

The UI is generated from live machine state and refreshed automatically by cron.

Business Surface

Projects

What matters commercially right now, with the million-dollar goal still staring at the wall like a polite threat.

Quantum Qlick

Primary revenue engine. AI consulting for businesses and the clearest lane toward the $1M target.

Signs Speak

Product in development. Needs continued clarification, scope discipline, and execution.

Habitrix

Product in development. Needs cleaner product-state tracking and sharper operating visibility.

Human skill stack

Chitundu

These are the skills currently associated with the human operator and should stay visible in the Command Center.

AI/techsalescontent creationvideo editingphotographychess coaching
Ops Notes

Notes

Operational notes with more substance and less decorative amnesia.

Current posture

This dashboard is generated from live machine state. Claude Code auth, Hermes posture, OpenClaw state, DeepSeek V4 registration, heartbeat cadence, human skills, Gravitas Lab status, and installed workspace skills should refresh automatically rather than relying on manual edits.

Change-control reminder

The user prefers richer operational information density. Major frontend information removal or oversimplification requires explicit user involvement before the change is made.