Push Claude Code beyond stock. Claude builds, Codex critiques. A gated research → plan → review → ship pipeline, with autonomous multi-agent revise loops that self-converge.
A Claude Code plugin built around a deliberate division of labor between two AI coding agents:
- Claude implements — planning, coding, subagents, agent teams
- Codex reviews — plan critique, code review, documentation accuracy review
Thesis: Claude is the builder, Codex is the critic. Better software with a smarter cost split.
┌─ refine ─┐ ┌──── fix ───┐ ┌──── fix ───┐
▼ │ ▼ │ ▼ │
research → plan → plan-review → implement → code-review → docs-sync → docs-review → ship
│ │ │ │ │ │ │ │
Codex+Claude Claude Codex Claude(+agents) Codex Claude Codex user
When the idea itself is vague (not just un-planned), an optional hyper-interview front-end clarifies it into a spec before research / plan — a short one-question-at-a-time interview, Claude-only (no Codex; clarity is its job, review happens downstream). The refine / fix arcs are what hyper-plan-loop, hyper-implement-loop, and hyper-docs-loop automate — a Claude-side teammate (planner / fixer / documenter) revises while Codex stays the reviewer, looping until no blocking findings remain. Gates write trace artifacts under .hyperclaude/ (gitignore-friendly); hyper-docs-sync, hyper-docs-loop, and hyper-implement edit the working tree directly. Skip any step a small change doesn't need — only code-review is non-negotiable for behavioral changes. See docs/workflow.md for triggers, skip rules, slug/artifact conventions, and --resume.
One gesture, end-to-end:
/hyperclaude:hyper-auto add OAuth login to the API
hyper-auto chains hyper-plan-loop → hyper-implement-loop. Claude plans, Codex critiques the plan until no blockers remain, Claude implements, Codex code-reviews until no blocking findings remain (style/nits are reported, never gating) — all hands-off. It's not a new layer, just composition over the two loops, so the same gates and artifacts apply. Requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 (inherited from the underlying loops).
User in Claude Code
│
┌────────────────┼───────────────┐
│ │ │
Commands Skills ────────► Agents
hyper-setup gates + orchestr. planner / implementer
(no spawn) │ verifier / documenter
▼ researcher / fixer
codex-bridge.mjs
(only Codex-spawning code;
always read-only sandbox)
│
▼
.hyperclaude/{specs,research,plans,epics,
plan-reviews,code-reviews,docs-reviews}/
Hooks — SessionStart reminder, fires independently
Four layers — Commands (explicit slash entry points), Skills (description-triggered gates, orchestrators, autonomous loops, and hyper-auto), Agents (Claude implementation arm), Hooks (SessionStart reminder). Every Codex spawn — fresh or --resume, with live web search enabled — runs in a read-only sandbox; Codex is critic, never editor. The bridge accepts two optional model-selection flags on all four modes: --model <name> (any non-empty model identifier, e.g. openai/gpt-5) and --effort <low|medium|high|xhigh> (reasoning effort; none and minimal are rejected); both default to null, inheriting ~/.codex/config.toml. code-review additionally accepts --background "<text>": a short, strictly descriptive change context (what the change is / what it touches / author intent) passed to the Codex critic to orient the review; it does not alter the review rubric. --background is optional and a no-op when omitted; it is mutually exclusive with --resume (resumed sessions already carry context in the Codex thread). The hyper-code-review skill composes and passes a background automatically; direct CLI callers may pass it explicitly. These are bridge-level flags; v1 is not exposed via slash commands. See docs/architecture.md for layer details, bridge internals, and the sandbox flag matrix.
External dependencies: Claude Code plugin runtime, codex-cli >= 0.130.0 with the global --search flag, Node 18+, and git. Nothing else (no npm bin, no tmux, no MCP servers).
-
Install:
/plugin marketplace add zeikar/hyperclaude /plugin install hyperclaude
-
Verify prerequisites:
/hyperclaude:hyper-setupChecks Node, codex-cli,
codex --search, git, and (optionally)CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS(needed by the loops andhyper-auto). Report-only; nothing is installed automatically. -
Run the cycle. Invoke gates explicitly, or chain them with the loops:
/hyperclaude:hyper-research add OAuth login to the API # Codex+Claude prior-art / pitfalls /hyperclaude:hyper-plan # Claude writes .hyperclaude/plans/<slug>.md /hyperclaude:hyper-plan-review # Codex critiques the plan /hyperclaude:hyper-implement # Claude executes the plan task-by-task /hyperclaude:hyper-code-review # Codex reviews the diff (branch vs main) /hyperclaude:hyper-docs-sync uncommitted # Claude updates docs for the change /hyperclaude:hyper-docs-review # Codex accuracy gate on docs # Or let the loops self-converge: /hyperclaude:hyper-plan-loop add OAuth login to the API # plan → review → revise, looped /hyperclaude:hyper-implement-loop <plan path> # implement → code-review → fix, looped /hyperclaude:hyper-docs-loop # docs → review → fix, looped (default: docs/) /hyperclaude:hyper-auto add OAuth login to the API # plan-loop → implement-loop, end-to-end
- docs/architecture.md — layers, bridge details, plugin layout, output contract.
- docs/gates-and-agents.md — what each skill and agent does, when to invoke.
- docs/workflow.md — the research → ship cycle, slug/artifact conventions, skip rules,
--resume. - docs/development.md — local install, tests, release flow.
- docs/decisions.md — non-obvious "why" notes and active deferrals.
node --test tests/*.mjs # unit tests for the bridge and setup-doctor
bash scripts/test/smoke.sh # acceptance smoke checksZero npm dependencies. Node 18+ stdlib only.
Structural inspiration from:
- superpowers by Jesse Vincent
- oh-my-claudecode by Yeachan Heo
No code ported from either; references only.