Official Compound Engineering plugin for Claude Code, Codex, Cursor, and more
22.9k
Stars
1.7k
Forks
86
Open issues
30
Contributors
AI Analysis
Compound Engineering is a Claude Code plugin system that operationalizes a structured development workflow (brainstorm → plan → work → simplify → review → compound) to reduce technical debt and increase development velocity. It is purpose-built for AI-assisted software engineering teams using Claude or similar LLM-based code editors, and is most valuable for teams already committed to planning-heavy, agent-centric workflows — not a general-purpose productivity tool for solo developers or team...
Inferred from signals mentioned in the README (tests, CI, type safety) — not a review of the actual code.
AI's overall editorial judgment — not an average of the bars above, can weigh other factors too.
Compound Engineering plugin brings structured AI dev workflows to Claude Code, Cursor, and Codex
Compound Engineering is a plugin for AI coding environments (Claude Code, Cursor, Codex) that imposes a structured workflow — brainstorm, plan, execute, review, document — intended to make each development cycle compound into better context for the next. Built by Every Inc, a company focused on AI-native productivity, it targets engineering teams using AI coding assistants who want to reduce repeated context loss and accumulating technical debt. With 21,811 stars and active maintenance as of June 2026, it has achieved meaningful visibility in the AI dev-tooling space.
Created in October 2025 by Every Inc, coinciding with the rapid expansion of plugin ecosystems for AI coding tools like Claude Code and Codex. Appears to be a formalization of workflows Every Inc developed internally and published on their editorial platform every.to.
The repo accumulated over 21,000 stars in roughly 8 months, likely driven by Every Inc's editorial reach (chain-of-thought newsletter/blog), the timing of Claude Code and Codex plugin marketplace launches, and organic sharing among developers frustrated with context loss in AI-assisted development. Recent weekly growth of 81 stars suggests the initial viral spike has subsided into steady, moderate traction.
Every Inc references internal use via published articles ('how Every codes with agents'), which provides some first-party adoption signal. External production usage beyond Every Inc's own team is not independently verified from available metadata. The 1,606 forks suggest meaningful engagement beyond casual starring, but verifiable third-party production deployment data is absent.
Likely a plugin package (npm: @every-env/compound-plugin) that ships slash-command definitions and agent configurations for multiple AI coding environments. Based on README, it appears to deliver 37 skills and 51 agents as composable units. Architecture is likely declarative — command definitions routed to AI models — rather than imperative code generation. Multi-environment support (Claude Code, Cursor, Codex) suggests adapter layers per platform.
A CI badge is present in the README linking to a GitHub Actions workflow, indicating automated testing exists. Specific coverage metrics or test methodology are not documented in the README.
Last push was June 20, 2026 — one day before the evaluation date — indicating very active maintenance. The presence of a CI pipeline, npm publishing, and multi-platform installation documentation all suggest an actively managed project, not a one-time release.
ADOPT IF: your team regularly uses Claude Code, Cursor, or Codex and suffers from lost context, inconsistent planning habits, or review debt in AI-assisted development cycles. AVOID IF: you want lightweight, low-ceremony AI assistance — the 80/20 planning-heavy philosophy adds real upfront friction that may not suit fast-moving solo projects or teams already with strong process discipline. MONITOR IF: you are evaluating structured AI workflow tooling but are uncertain whether the plugin ecosystem for your preferred tool (especially Codex's partial support noted in README) is mature enough for reliable daily use.
Independent dimensions
Mainstream potential
5/10
Technical importance
7/10
Adoption evidence
3/10
- Platform dependency: the plugin is tightly coupled to Claude Code, Cursor, and Codex plugin APIs. Breaking changes or deprecations in any of these platforms could require significant rework.
- Workflow overhead risk: the prescribed brainstorm-plan-review-compound cycle may feel bureaucratic in smaller or faster-moving projects, potentially reducing rather than increasing velocity.
- Codex integration is noted in the README as partially manual (TUI install required, CLI limitation acknowledged), suggesting the multi-platform story is not uniformly polished.
- Adoption concentration: growth appears closely tied to Every Inc's editorial platform and audience. Without broader community contribution, the project may remain dependent on a single organization's priorities.
- Verification gap: the core claim — that compounding documentation genuinely improves future AI context — is philosophically sound but not independently benchmarked in available materials.
Likely to maintain steady niche adoption among AI-native engineering teams, particularly those aligned with Every Inc's workflows. Broader mainstream adoption will depend on whether plugin ecosystems for Claude Code and Codex mature and whether the structured-workflow philosophy gains traction beyond early adopters.
Newsletter
Get analyses like this every Monday
Free weekly digest of the most interesting open-source discoveries.
Languages
Information
- Language
- TypeScript
- License
- MIT
- Last updated
- 9h ago
- Created
- 9mo ago
- Analyzed with
- anthropic/claude-haiku-4-5
Stars over time
Contributors over time
Top 100 contributors only — repos with more will plateau at 100.
Open issues
Codex tool map collapses subagent dispatch to the main thread, silently breaking skills that require isolated executors
ce-compound: Discoverability Check silently edits AGENTS.md/CLAUDE.md in headless (skill-to-skill) mode
Support configurable work delegates beyond Current Coding Agent and Codex
fix: bun test fails on main - release-metadata skill count is stale (28 -> 29)
test-permission-probe
Open pull requests
Top contributors
Similar repos
notlikeDev/CCPlugins
CCPlugins is a collection of 24 professional commands extending Claude Code CLI...
YishenTu/claudian
Claudian is an Obsidian plugin that integrates Claude Code, Codex, and other AI...
coleam00/context-engineering-intro
A template and methodology for 'context engineering'—a structured approach to...
NeoLabHQ/context-engineering-kit
Context Engineering Kit is a curated collection of TypeScript-based prompt...
| Repository | Stars | Week Δ | Language | Score | Updated |
|---|---|---|---|---|---|
|
EveryInc/compound-engineering-plugin
this
|
22.9k | +416 | TypeScript | 8/10 | 9h ago |
|
notlikeDev/CCPlugins
|
2.7k | — | Python | 7/10 | 2w ago |
|
YishenTu/claudian
|
13.8k | — | TypeScript | 8/10 | 6h ago |
|
coleam00/context-engineering-intro
|
13.7k | — | Python | 7/10 | 4mo ago |
|
NeoLabHQ/context-engineering-kit
|
1.2k | — | TypeScript | 8/10 | 4d ago |
 openai/codex-plugin-cc
|
27.2k | — | JavaScript | 7/10 | 2d ago |
Focuses on context engineering as a concept with instructional content (13,479 stars, Python). Less structured as an installable plugin; more of a reference/tutorial. Compound Engineering is more opinionated and workflow-complete by comparison.
OpenAI's official Codex plugin (21,349 stars, JavaScript) likely targets raw Codex integration rather than imposing a workflow methodology. Compound Engineering adds an opinionated layer on top of such integrations rather than competing directly.
A TypeScript project for Claude-based agents with similar star counts (12,972). Appears to be a different angle on Claude tooling — exact overlap is unclear from available metadata, but both serve developers extending Claude Code workflows.
Microsoft's offering (2,772 stars, TypeScript) is likely more enterprise-focused and narrower in scope. Compound Engineering appears more developer-facing and immediately installable via plugin marketplaces.
General OpenAI plugins reference repository (3,262 stars). Foundational infrastructure rather than a workflow methodology — serves a different purpose and is not a direct alternative.
