Warp Speed Configs

Configuring a Repo for High Warp

Warp is delegation intensity — how much autonomy you hand your agents. Everyone flies low warp; you earn high warp by wiring the repo so delegation is safe. This series walks the levels one at a time, then re-cuts the same material by feature so you can configure one capability end-to-end.

This is the user-facing configuration guide. The internal licensing and credential mechanics — how warp ratings are earned and enforced — are separate and internal-only; nothing here depends on them.

The climb, at a glance

Four configuration levels. Each one earns the next: you don't jump to autopilot, you wire the safety and bookkeeping that make autopilot survivable. Climbing warp takes weeks, not days — you and the agent agree a cadence (a couple of items a day) and hit KPIs as you go.

The “forward-deployed agent” model

We don't ship a wizard and we don't rely on hooks. We ship an agent that deploys into your repo like an embedded engineer and levels it up with you over weeks. Instruction delivery is command-driven, not hook-driven — more reliable and less intrusive: no per-machine settings.json wiring, and everything the agent does lands as durable, reviewable files in your repo.

The mechanism is two parts — a one-time bootstrap prompt, and an npm client the agent drives:

# the agent runs these for you, as commands — not a hidden hook
cloudclawer-warp install         # install the client into the repo
cloudclawer-warp add-roles       # write role files (requires an active subscription)
cloudclawer-warp wire-claude-md  # insert a managed section into CLAUDE.md
cloudclawer-warp update          # pull latest roles + instructions, brief you on changes

Because the instructions live in CLAUDE.md (which the agent reads every session) and the roles live as files in the repo, the agent continues day after day with no hook required. The managed section itself tells the agent to run update at the start of a session and brief you — that is the auto-update loop, driven by the agent, not a hook.

New to the commands? The Forward-Deployed Agent feature page walks each one — what it writes, when to run it, and how the managed CLAUDE.md markers work.

Two ways to read this series

Same material, two cuts. Read by level if you're onboarding a repo and want a checklist to climb. Read by feature if you already know which capability you're configuring and want it end-to-end.

Warp 1

Connected

Get the agent talking to your tools.

Warp 2

Assisted & Auditable

Reversible work autonomous; a human gate on prod.

Warp 3

Delegated & Specialized

Multiple agents, roles, and the bookkeeping to justify spend.

Warp 4

Autopilot

Routines launch work themselves while you're away.

Or jump to a feature

The Forward-Deployed Agent

The embedded-engineer model and the npm client commands that drive it.

Repo Hygiene

CLAUDE.md and an organized issue tracker — the substrate everything else needs.

GitHub, CI & Auto-Merge

Push access, auto-merge after CI, and PR/pipeline visibility.

Autonomous Workflows

Walking each workflow and giving the sandbox the tools it needs.

Credentials & the Safety Spine

GitHub Actions as the vault, indirect access, and data-loss guards.

Experiment Runners

Local experiment runs as a third deploy target — workflow_dispatch onto your own hardware, results read back.

Roles & Orchestration

Role files and foreground subagents under the Chief of Agents.

Bookkeeping & Cost

Cost per ticket, tabulated — the only way to judge if spend is worth it.

Routines & Tech Tree

Prioritization routines, sprint cadence, and the dependency tech tree.

The Chief of Agents' standing role

When you run CloudClawer, the Chief of Agents (CoA) takes care of your needs and CloudClawer manages the instructions for you. The CoA leaves notes about your repo and the outstanding items needed to push warp higher, sets an agreed schedule with you, and runs scheduled maintenance — which may include security reviews. If you opt in to update checks, these help keep your agent instructions current, and the agent briefs you when an update is present.

High warp is earned, not toggled. Every level assumes the one below it is already wired — skip the safety spine at Warp 2 and you'll be granting autonomy you can't revoke. Start at Warp 1.