wee.cat

The Problem

AI can write code.
But where does it run?

Agents need real environments — filesystems, shells, git, toolchains. Running them locally is risky. Setting up VMs takes forever. Chat windows aren't enough.

Local is risky

Your files, credentials, and system exposed to autonomous agents.

Setup takes forever

Docker, VMs, cloud instances — hours of config before an agent writes a line.

No visibility

Most setups give you no way to watch what the agent is actually doing.

How It Works

Your agents. Your repos.
In parallel.

Multi-provider agents

Bring Claude, GPT, Gemini, or open-source models. Wee is the runtime, not the model.

Work across projects

Spin up sandboxes for different repos simultaneously. Each fully isolated.

Safe with git worktrees

Agents work on branches in parallel — no conflicts, no stepping on each other. Just results.

GPU On Demand

Need a GPU?
Just ask.

Agents can spin up powerful RunPod GPUs as a sidecar — RTX 4090, A100, H100, even B200 — directly from the sandbox via MCP. Train models, run inference, then shut it down. Pay only for what you use.

Powerful GPUs on tap

RTX 4090, A100, H100, up to B200. Auto-fallback if a tier is sold out.

Agent-controlled via MCP

Launch, push code, run commands, pull results — all through MCP tools the agent already knows.

Seamless sidecar

The GPU pod lives alongside your sandbox. rsync files in, SSH to run, pull results back. Zero config.