skill

clawvis

ldom1/clawviscommunitylistedunreviewed

Set up

Skills are installed by placing their folder where your agent loads skills.

Download ZIP from GitHub Open the repository

The download comes from github.com/ldom1/clawvis itself, not from our server, so what you get is exactly what its author published.

Install instructions come from the README

Skills do not use an MCP config. Follow the steps in the README below.

README

synced Oct 22, 2026

One-liner install: curl -fsSL https://raw.githubusercontent.com/ldom1/clawvis/main/get.sh | bash

Install

One command — no setup required:

curl -fsSL https://raw.githubusercontent.com/ldom1/clawvis/main/get.sh | bash

Clones the repo to ~/.clawvis, runs the interactive wizard, and starts the stack.

With git (control where it lives):

git clone https://github.com/ldom1/clawvis && cd clawvis && ./install.sh

Non-interactive (CI / scripted):

./install.sh --non-interactive --instance myname --mode docker

After install, open http://localhost:8088 and configure your AI runtime in Settings → AI Runtime.

What is Clawvis?

You run AI agents — Claude, Mistral, or your own OpenClaw instance. But you have no single place to see what they're doing, manage their tasks, or keep project notes alongside them.

Clawvis is that place. One docker compose up to rull them all.

Service	URL	What it does
Hub	`localhost:8088`	Dashboard — system status, agent activity, projects
Agent service	proxied as `/api/hub/agent/*`	Streaming chat and runtime config for the Hub banner
Kanban	`localhost:8088/kanban/`	Task board with confidence scoring and memory sync
Brain	`localhost:8088/memory/`	Project knowledge base (markdown → searchable pages)
Logs	`localhost:8088/logs/`	Real-time log stream from all your agents
Settings	`localhost:8088/settings/`	AI runtime config, workspace paths, linked instances

Your data stays in instances/<your-name>/ — never touched by core updates.

Projects on the Hub home support single delete from the card menu and bulk delete by selecting multiple cards then clicking Delete selected.

Prerequisites

Mode Franc — default (Docker)

Everything runs inside containers. You only need:

Tool	Why	Install
Docker Engine 24+ or Desktop	runs the full stack	docs.docker.com
Git	clones the repo (`get.sh` handles this)	pre-installed on most systems

That's it.

Mode Soissons — dev / contribution

Running the stack locally (outside Docker) requires:

Tool	Version	Why
Node.js	>= 18	Vite dev server + CLI wizard
npm	bundled with Node	CLI dependencies
Yarn (via corepack)	4.x	Hub frontend build (`corepack enable` installs it automatically)
uv	latest	Kanban API + hub-core — `curl -LsSf https://astral.sh/uv/install.sh \| sh`

Mode Mérovingien — VPS / server deploy

Choose 2) Mérovingien in the wizard for deployments where nginx or another reverse-proxy is already running on the target port. The installer creates the instance structure and .env without launching Docker, so you keep full control of startup:

# Non-interactive equivalent:
./install.sh --non-interactive --instance myname --mode docker --no-start

# Then start manually (with instance override):
docker compose \
  -f docker-compose.yml \
  -f instances/myname/docker-compose.override.yml \
  up -d

Daily use

clawvis start       # start the stack (dev mode)
clawvis doctor      # health check — shows what's up and what's not
clawvis shutdown    # graceful stop
clawvis restart     # stop + start

When clawvis is installed globally (~/.local/bin/clawvis) and you run commands inside another Clawvis checkout, start/restart now use the current checkout so local updates are applied.

Cron + Telegram: docker compose up (and clawvis restart in Docker mode) start telegram and scheduler with the rest of the stack. Without TELEGRAM_BOT_TOKEN, the telegram service runs in stub mode (HTTP /send only — logs deliveries, no Telegram API). Set TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID in .env for a real bot. TELEGRAM_CHAT_ID must be the target chat/channel id (not the bot id), otherwise /test and scheduler notifications are rejected. Cron definitions: services/scheduler/definitions/jobs/*.yaml and workflows in services/scheduler/definitions/workflows/*.yaml.

Telegram memory: The bot remembers across sessions. When BRAIN_PATH is set and docker-compose.clawvis-brain.yml is active, the telegram service reads a lightweight memory context (user profile + active context + open topics, ≤300 tokens) at startup and injects it into every prompt. On shutdown it asks the agent to extract key points and writes them back to resources/knowledge/operational/clawvis/ in the brain vault — no conversation transcripts, just distilled facts. brain-sync commits and pushes the updated files at session end.

AI Runtime

No API key is required at install time. Configure post-install:

Option A — Browser (recommended): Go to http://localhost:8088/settings/ → AI Runtime section. Keys are stored in your browser's localStorage.

Option B — .env file (backend / persistent): Edit .env and restart:

CLAUDE_API_KEY=sk-ant-...        # Anthropic Claude
MISTRAL_API_KEY=...              # Mistral
OPENCLAW_BASE_URL=http://host    # Self-hosted OpenClaw
OPENCLAW_API_KEY=...             # OpenClaw key (optional)

Stay up to date

Core updates never touch your instance data:

clawvis update status
clawvis update --tag v2026-03-23
clawvis update --channel stable

Backup and restore:

clawvis backup create
clawvis backup list
clawvis restore <backup-id>

Deploy to a server (Hostinger / VPS)

1. Set deploy target in .env:

DEPLOY_HOST=your-vps-ip
DEPLOY_USER=ubuntu
DEPLOY_PATH=/opt/clawvis
DEPLOY_SSH_PORT=22

2. Deploy:

clawvis deploy

See docs/guides/deploy-hostinger.md for the full guide.

Brain (Knowledge base)

The Brain shows your project markdown files from instances/<name>/memory/projects/.

On the Hub home, each project that is not the home “active” row has an Activate project button (and a ⋮ menu for archive/delete). It sets YAML frontmatter status: active on that project’s memory note so the card moves to the main grid (others stay under “Show all projects” until activated).

Claude Code in Docker: the Kanban API mounts your host ~/.claude and sets CLAWVIS_HOST_CLAUDE_DIR / CLAWVIS_REPO_HOST_PATH (see docker-compose.yml) so runtime setup updates the same files Claude Code reads on the machine, not paths inside the container. claude.json points node at <repo>/mcp/server.js on the host (not /clawvis/...). One-time: cd mcp && npm install so the MCP SDK is available, then claude refresh. Optionally mount the host claude binary and set CLAWVIS_HOST_CLAUDE_CLI for wizard detection.

It works out of the box — no Quartz installation needed. A lightweight Python renderer converts your .md files to HTML automatically. If you want the full Quartz static site experience, clone Quartz to quartz/ in the repo root and run clawvis start.

Launching a project on dombot.tech

Projects created in Clawvis can be exposed at https://<project>.dombot.tech via the clawvis-deployment Ansible repo.

How it works

Browser (HTTPS)
  → VPS nginx — per-project Let's Encrypt cert
    → Tailscale tunnel
      → devbox nginx — routes by subdomain
        → container on 127.0.0.1:<port>

Each project gets its own subdomain, its own TLS certificate, and a port in the 8100–8199 range on devbox.

Register a new project subdomain

From the clawvis-deployment repo:

uv run ansible-playbook add_project.yml \
  -e project_name=myapp \
  -e project_port=8100

One command provisions everything: nginx vhosts on both devbox and VPS, a Let's Encrypt cert via HTTP challenge, and HTTP→HTTPS redirect. The project is live at https://myapp.dombot.tech within seconds.

See docs/routing-architecture.md for the full architecture.

Remove a project subdomain

uv run ansible-playbook remove_project.yml -e project_name=myapp

Port allocation

Range	Usage
8088	Clawvis hub (fixed)
8090–8092	Clawvis internal services (fixed)
8100–8199	Clawvis projects (dynamic)

What's inside

Directory	Purpose
`hub/`	Vite SPA frontend + nginx Docker image
`hub-core/`	Python lib — identity, RBAC, AI adapters
`services/kanban/`	Task board FastAPI — tasks, projects, memory sync
`services/agent/`	AI agent service — streaming chat, runtime config
`services/scheduler/`	Cron job runner
`services/telegram/`	Telegram bot
`skills/`	Pre-configured agent skills (kanban, logger, brain…)
`skills/project-init/templates/`	Starter templates for new projects
`instances/`	Your instance data — never overwritten by updates

Troubleshooting

Hub shows blank page: The Kanban API may not be running. Check: clawvis doctor or docker compose ps.

clawvis: command not found after install: Reload your shell: source ~/.zshrc (or ~/.bashrc).

Docker not running: Start Docker Desktop, or: sudo systemctl start docker.

Port already in use: Change HUB_PORT in .env (e.g. to 8089), then restart.

AI runtime shows "not configured": Go to localhost:8088/settings/ and configure your provider.

Contributing

Contributions welcome! See CONTRIBUTING.md. Mode Soissons (contribution) in the installer sets up your dev environment.

./install.sh  # choose mode: dev
clawvis start

License

MIT