Bob Ross
Manages a Canonical Landscape estate, providing inventory, alerts, activities, patching, and script execution with a safety layer including blast-radius preview, dry-run confirmation handshake, and full audit logging.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Bob Rossshow computers with pending security updates"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Bob Ross ๐จ
"We don't make mistakes, just happy little servers."
An MCP server for managing a Canonical Landscape estate from an AI agent (Claude Code / Desktop). Bob Ross wraps the Landscape API as calm, friendly MCP tools โ inventory, alerts, activities, patching, script execution, reboots โ with a safety layer built right in, so your agent can tend a whole fleet of happy little servers without ever beating the devil out of the wrong one.
Everybody needs a friend. Even your Ubuntu estate.
๐๏ธ Why it's not just a dumb API wrapper
A blank canvas is a beautiful thing โ and so is a fleet that's still standing. Bob Ross paints carefully:
Blast-radius preview โ before you touch anything,
resolve_querysteps back from the easel and shows you how many machines a query matches (and a sample). Always know how big the canvas is.Dry-run โ confirm handshake โ every destructive tool returns a short-lived
confirm_tokenon the first call. You have to call again with the token to actually do it. If the set of matched machines drifts in between, the token is refused. No happy little accident reboots 500 boxes.Secure by default โ read-only mode is on out of the box (we all start with a clean canvas). Writes need two switches flipped on purpose. TLS is verified. Secrets never touch the logs.
Full audit log โ every stroke (dry runs, executes, denials) is appended to a redacted JSONL trail. You can always see what the brush did.
Activity-aware โ write actions in Landscape run later, asynchronously. Pass
wait=trueand Bob Ross watches the paint dry, then tells you succeeded / failed / still-going per machine โ not just "queued."Dual auth โ legacy HMAC query API or REST bearer token, auto-detected.
Related MCP server: MCP Process Server
๐จ Get the paints out (install)
From PyPI โ the whole studio in one command:
pip install bob-ross-landscapeThat gives you the bob-ross command (the import package is bob_ross).
git clone https://github.com/just-an-oldsalt/bob-ross
cd bob-ross
python3 -m venv .venv && source .venv/bin/activate
pip install -e ".[dev]"
pytest # unit tests (safety / signing / config / health / ...)
python scripts/smoke.py # safe live smoke test vs your instance (read-only, no writes)๐ผ๏ธ Set up your palette (configuration)
Everything is BOBROSS_* env vars (or a .env file โ see .env.example):
Var | Default | Meaning |
| โ | Landscape base URL (no trailing |
| โ | Legacy HMAC keys (mode A) |
| โ | REST bearer token (mode B) |
|
| Blocks all writes โ the safe blank canvas |
|
| The second switch needed to enable writes |
|
| Verify TLS certs (set |
To let Bob Ross pick up a brush, flip both
BOBROSS_READ_ONLY=falseandBOBROSS_ALLOW_WRITES=true. Even then, every destructive action still needs the dry-run โ confirm handshake. Belt and suspenders. Happy and safe.
๐ช Add it to Claude Code
pip install bob-ross-landscape
claude mcp add bob-ross --scope user \
--env BOBROSS_LANDSCAPE_URL=https://landscape.example.com \
--env BOBROSS_ACCESS_KEY=YOUR_KEY \
--env BOBROSS_SECRET_KEY=YOUR_SECRET \
-- bob-rossPrefer to keep secrets out of Claude's config? Put them in a .env file and run
from that directory instead:
claude mcp add bob-ross --scope user -- \
bash -lc 'cd /path/to/your/bobross-env && exec bob-ross'Claude Desktop (stdio)
{
"mcpServers": {
"bob-ross": {
"command": "bob-ross",
"env": {
"BOBROSS_LANDSCAPE_URL": "https://landscape.example.com",
"BOBROSS_ACCESS_KEY": "YOUR_KEY",
"BOBROSS_SECRET_KEY": "YOUR_SECRET"
}
}
}
}Bob Ross always wakes up in read-only mode. Start there. Get comfortable. Then, when you're ready, let's get a little crazy.
๐ฒ The brushes (tools)
Read (always safe โ look all you like):
ping ยท estate_health ยท list_computers ยท get_computer ยท resolve_query ยท
pending_updates ยท list_alerts ยท list_activities ยท get_activity ยท
wait_for_activity ยท list_scripts
Write (gated โ dry-run โ confirm every time):
execute_script ยท reboot_computers ยท apply_security_upgrades ยท
upgrade_packages ยท install_packages ยท remove_packages ยท add_tags ยท
remove_tags
Activity-creating write tools take
wait=trueto poll the resulting Landscape activity to a terminal status and hand back acompletionsummary (succeeded / failed / still-incomplete) โ so your agent knows the real outcome, not just "queued."
Resources: landscape://computers ยท landscape://alerts ยท
landscape://health ยท landscape://computer/{computer_id} (template)
Prompts: patch_security_updates ยท triage_estate ยท
reboot_reboot_required ยท patch_machine
๐๏ธ A happy little workflow
You have unlimited power here. Move mountains โ one confirmed step at a time:
you: "what needs my attention across the fleet?"
โ estate_health โ "kaylee-mc: 138 pending upgrades, 3 boxes need reboots"
you: "what would patching kaylee-mc actually change?"
โ pending_updates title:kaylee-mc โ the list, per package
you: "apply the security upgrades there"
โ apply_security_upgrades โ dry-run shows blast radius + a confirm_token
โ (you approve) โ re-run with the token + wait=true
โ completion: succeeded โ
no failed patches, just happy little servers๐ฆ Publishing & links
Releases: tag
vX.Y.Z,gh release createโ GitHub Actions publishes to PyPI via OIDC (no tokens). SeePUBLISHING.md.MCP Registry manifest:
server.json
"Talent is a pursued interest. Anything you're willing to practice, you can do."
Now go tend some happy little servers. ๐จ
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/just-an-oldsalt/bob-ross'
If you have feedback or need assistance with the MCP directory API, please join our Discord server