spore.host
Open Source · Apache 2.0

Compute that manages itself

Find the right EC2 instance. Launch it in minutes. Auto-terminate when your job is done. Built for researchers and data scientists, not ops engineers. It runs on your own AWS account — there's no new compute provider to sign up for, and you pay AWS only for what you use.

Avoid surprise bills GPU-aware Slack & Teams control AI assistant integration macOS, Linux, Windows
Launch your first protected instance See how it works Browse source

Runs in your AWS account · Credentials stay with you · Hard TTL on every protected instance · Open source · No standing cluster

AMD EPYC Genoa search · launch with TTL + idle timeout · run a job · check cost · watch it self-terminate

Start from your goal

What are you trying to do?

Start a temporary research workstationspawn → connect Run a job and terminate when it's donespawn task / job Find the right CPU or GPU (and its price)truffle Run many independent jobs at oncearrays / sweeps Wait for scarce GPU capacity to appearlagotto Use an existing workflow engineNextflow / Snakemake / CWL / WDL / Airflow

The tool suite

Six tools. One workflow.

Each tool has exactly one job. Together they cover the full lifecycle of ephemeral compute.

Discovery

Truffle

Find the right instance before you commit. Search by GPU model, vCPU count, memory, or plain language. Compares Spot prices across regions and checks your service quotas.

  • Natural language search — "nvidia h100 8gpu"
  • Spot prices across all regions and AZs
  • Quota check before you launch
  • On-Demand Capacity Reservations

Docs · Source · Releases

Launch & manage

Spawn

Launch EC2 instances in under two minutes. Interactive wizard, CLI flags, or pipe from truffle. Instances are tagged with a TTL and auto-terminate — no cron jobs, no forgotten instances.

  • TTL auto-termination
  • Idle detection — CPU, network, processes, sessions
  • Hibernation support
  • Spot, GPU, MPI clusters, parameter sweeps

Docs · Source · Releases

Lifecycle daemon

Spored

The small agent that runs on every launched instance and enforces its lifecycle from the inside — so nothing depends on your laptop staying open. You never install it: spawn provisions it automatically.

  • Enforces the TTL hard deadline
  • Idle detection — CPU, network, processes, sessions
  • Fires completion, TTL-warning, and idle events
  • Runs pre-stop hooks to save work before shutdown

Docs · Source · Releases

Capacity watching

Lagotto

Watch for instance availability and act when it appears. Lagotto polls for capacity across regions on a schedule — useful for scarce GPU types where you wait for a spot to open up.

  • Polls across regions on configurable schedule
  • Triggers launch, notification, or webhook
  • Serverless — runs as a Lambda function
  • Slack notifications when capacity appears

Docs · Source · Releases

Chat control

Spore-bot

Control your instances from Slack or Microsoft Teams without opening a terminal. Start, stop, check status, extend runtime, and receive lifecycle notifications as direct messages.

  • /spore status, start, stop, extend
  • DM notifications on idle, TTL warning, completion
  • OAuth — one-click workspace connection
  • Works from any device, anywhere

Docs · Source · Setup

AI assistant

MCP Server

Works with any MCP-compatible AI assistant — Claude Code, Claude Desktop, Cursor, Windsurf, and others. Find instances, check status, and manage compute through natural conversation.

  • Works with Claude Code, Claude Desktop, Cursor, and more
  • Natural language instance search via truffle
  • Manage running instances via spawn (status, stop, terminate, extend — no launch, by design)
  • Install via Homebrew — one config line

Docs · Source · Releases

Installation

Up and running in minutes

Install the tools that fit your workflow. Truffle and spawn are the core pair. Add the others as needed.

# Core tools
brew install spore-host/tap/truffle
brew install spore-host/tap/spawn

# Optional: capacity watcher
brew install spore-host/tap/lagotto

# Optional: MCP server for AI assistants
brew install spore-host/tap/spore-host-mcp
scoop bucket add spore-host https://github.com/spore-host/scoop-bucket
scoop install truffle spawn

# Optional
scoop install lagotto spore-host-mcp
# Assets use lowercase GOOS/GOARCH (darwin|linux, amd64|arm64),
# so normalize uname before matching the release asset.
OS=$(uname -s | tr '[:upper:]' '[:lower:]')
case "$(uname -m)" in
  x86_64|amd64) ARCH=amd64 ;;
  arm64|aarch64) ARCH=arm64 ;;
esac
for tool in spawn truffle; do
  url=$(curl -fsSL "https://api.github.com/repos/spore-host/${tool}/releases/latest" \
    | grep -o "https://[^\"]*_${OS}_${ARCH}.tar.gz" | head -1)
  curl -fsSL "$url" -o "${tool}.tar.gz"
  tar -xzf "${tool}.tar.gz" "$tool"
  sudo install "$tool" /usr/local/bin/
done

See other installation methods in the docs.

Debian/Ubuntu and RHEL/Fedora packages (.deb/.rpm) are also available on the releases page.

AI assistant integration

Ask your AI assistant to manage your compute

spore-host-mcp is a standard MCP server — it works with any MCP-compatible AI tool: Claude Code, Claude Desktop, Cursor, Windsurf, and others. Ask it to find instances, check status, or stop a forgotten job in plain English.