Colab Runner — Development agent for Claude Code
Executes GPU work on a Colab runtime from this terminal with the Colab CLI — provisions the accelerator, clones the branch, launches the run detached, watches it, and brings the artifacts back.
How to install Colab Runner
Installs to ~/.claude/agents/evolvingagentslabs-lora-kernel-colab-runner.md
mkdir -p ~/.claude/agents && curl -fsSL https://raw.githubusercontent.com/EvolvingAgentsLabs/lora-kernel/HEAD/.claude/agents/colab-runner.md -o ~/.claude/agents/evolvingagentslabs-lora-kernel-colab-runner.md Restart Claude Code, or start a new session, for it to be picked up.
What Colab Runner does
name: colab-runner description: Executes GPU work on a Colab runtime from this terminal with the Colab CLI — provisions the accelerator, clones the branch, launches the run detached, watches it, and brings the artifacts back. Use for any step in docs/PLAN.md that trains or serves an adapter. tools: Read, Bash, Grep, Glob, Edit model: sonnet
You run the steps this machine cannot. It is a 16 GB arm64 Mac: it cannot serve vLLM and cannot hold a 26B, and letting that shrink the experiment i
Alternatives in Development
- Rust Async Patterns — Production patterns for async Rust programming with Tokio runtime, including tasks, channels, stream 31.9k ★
- Accessibility Runtime Tester.Agent 31k ★
- Glossary — Quick reference for technical terms used throughout this course 2.2k ★
Full documentation available on GitHub
View Source RepositoryRelated Agents
Ndv Telemetry
Observability specialist. Use when adding logging, metrics, distributed tracing, or health checks. Detached ob
Docs Editor
Edits documentation in two modes. Drift mode takes a diff and a list of stale pages, as docs-reviewer names th
Rust Perf Engineer
Reviews diffs touching poly's per-file runner / discovery / cache / engine paths for hot-path regressions — ne
Xcode Runner
Runs Xcode apps on simulators, physical devices, or macOS. Determines if a rebuild is needed, delegates builds
Outcome Predictor
Independent reviewer for reform analyses. Mode 1 (predict) reads ONLY the reform provisions and predicts expec
Configuration
Stage 4 of Vulpine. Given the source tree and the codenav index, produce configure-target.sh — a bash script t
Related Skills
Restart All
Rebuild and restart the whole local agent-runner fleet via scripts/build_and_restart_all.py (self-last, detach
Claude n8n Bridge
Self-hosted n8n in Docker wired to Claude Code: n8n owns schedules and webhooks, a host-side runner executes o
Omk
Evidence-gated runner for Codex, Claude Code, OpenCode, and local coding agents. Routes tasks into scoped DAG