AIWG
Reusable project context and specialist workflows for the AI coding tools you already use, so decisions persist across sessions.
- Source repo
- jmagly/aiwg
- Stars
- ★ 225
- Last updated
- today
- License
- MIT
- Primary language
- TypeScript
- FA score
- Insufficient evidence
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude CodeClaude API (Unsupported)
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- An engineering team starting a multi-month feature project that needs architecture, security, and testability reviews with explicit phase-gate approvals.
- Not a fit if
- One-off tasks needing no saved context or review
- Teams unwilling to maintain .aiwg/ artifacts or configure providers
- Source review
- Insufficient evidence 8 safety controls not found
What does this agent do, and when should you use it?
AIWG is a cognitive-architecture framework for AI-augmented software development that packages agents, skills, rules, templates, and workflow utilities into a single source of truth deployable across 15 named AI platforms. Its deployment layer copies Markdown/YAML source files into the provider-native paths each tool reads via aiwg use, while project outputs persist in the .aiwg/ artifact directory for reuse across sessions. Six core components cover semantic memory, multi-agent deliberation with synthesis, the Ralph closed-loop self-correction cycle, bidirectional traceability, Cooper Stage-Gate phase planning, and controllable voice generation. Nine domain frameworks (SDLC Complete, Forensics Complete, Research Complete, and others) ship alongside dozens of opt-in addons such as Agent Loop, Compound Memory, and Dataset Intelligence. It is not a model and does not replace a provider subscription: the deployment core writes plain-text files the platform reads natively, and optional orchestration components must be explicitly enabled.
After installing, aiwg use all --provider <provider> copies kernel skills, specialist agents (Security Auditor, Test Architect, etc.), rules, and templates into provider-native directories (.claude/skills/, .codex/agents/, and so on), builds an artifact index, emits AIWG.md / AGENTS.md at the project root, self-verifies the deployment, and reports whether a provider reload is required. Day to day, you describe outcomes in natural language; the assistant searches AIWG's capability graphs, loads matching skills, and follows workflows shaped as Primary Author → Parallel Reviewers → Synthesizer → Human Gate → Archive. Work products — requirements, architecture decisions, test strategies, deployment plans — are saved in .aiwg/, cross-referenced with @-mentions, and linked from code via @implements annotations for doc-to-code-to-test traceability. The optional Ralph addon runs bounded execute-verify-learn-retry iterations with checkpoints, while aiwg doctor diagnoses deployment health.
- An engineering team starting a multi-month feature project that needs architecture, security, and testability reviews with explicit phase-gate approvals.
- A legacy system migration requiring phased rollback strategies and architecture decision records that survive across sessions.
- Compliance-heavy domains (healthcare, finance, aerospace) that need audit trails and a traceable requirements-code-tests chain.
- An incident-response team using the forensics framework for evidence acquisition, log analysis, persistence hunting, timeline building, and IOC extraction.
- Marketing or research teams that need reusable briefs, source records, citation checking, and content that carries forward between sessions.
- An individual developer who wants structured reviews inside an existing Claude Code, Codex, or Cursor session without switching tools.
How do you install or deploy this agent?
Prerequisites: Node.js >= 20.0.0 (prefer Node 24 for new installs) and a supported AI platform. Standard install: npm i -g aiwg, then cd /path/to/your/project && aiwg use all --provider claude (replace claude with your provider: codex, copilot, cursor, warp, factory, opencode, devin, and others). For a lighter install resolving signed, versioned resources, use npm i -g @aiwg/cli. Optional: install AIWG Cockpit with the self-hosted Agentic Sandbox executor (Docker or KVM/libvirt) via the guided prompt at https://aiwg.io/agentic-sandbox/setup.aiwg.yaml. On macOS, if npm fails with EACCES under /usr/local/lib/node_modules, follow the macOS install guide and avoid sudo global installs.
How do you use this agent?
After deployment, issue natural-language requests in your AI tool, such as "Set up the SDLC framework for this project and my AI tool" or "Review this design for security risks" — the assistant handles discovery, artifact lookup, and workflow tools. Deploy individual frameworks on demand: aiwg use sdlc --provider claude, aiwg use forensics --provider claude, aiwg use research --provider claude, etc. Maintenance commands: aiwg refresh regenerates context, aiwg doctor checks deployment health, and npx aiwg doctor works without PATH changes. Author project-local extensions directly under .aiwg/{extensions,addons,frameworks}/<name>/ for automatic discovery by aiwg use, no fork required.
What are this agent's strengths and limitations?
- One source of truth deploys to 15 named platforms plus a generic adapter, eliminating per-tool duplication of instructions.
- The
.aiwg/artifact memory plus bidirectional traceability (Doc↔Code↔Tests) makes decisions and review findings inspectable and auditable across sessions. - Unusual breadth: SDLC Complete alone ships 100 agents, 116 skills, 217 templates, and 39 rules, with eight more domain frameworks including forensics, marketing, research, ops, and security engineering.
- Project-local extensions are byte-identical to their upstream form, enabling zero-rewrite, hash-verified promotion.
- Requires Node.js >= 20 and an existing AI platform subscription; it replaces neither, and most workflows operate on project artifacts rather than an independent runtime.
- Installation has many documented failure paths: three deployment scopes, optional native features, macOS EACCES, and PATH issues.
- The README states plainly that a role definition is not a separate model and a written rule is not proof of enforcement; multiple reviewers can share an error, and benefits depend on keeping artifacts current.
- A full lifecycle is overkill for small tasks; extra context, reviewers, and verification steps add model calls and human effort.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AIWG This agent | Insufficient evidence | CLIFree + model costs | ★ 225 | today | TypeScript | Codex · Claude Code |
| Loop Library: Reusable AI Agent Workflows | 61 · Some gaps | Agent plugin / skillFree | ★ 3.2k | 3mo ago | JavaScript | Codex · Claude Code |
| CoCo Super Intelligence | 69 · Some gaps | Agent plugin / skillFree + model costs | ★ 785 | 1d ago | Python | Codex · Claude Code |
| OSpec | 50 · Major gaps | CLIFree + model costs | ★ 391 | 1mo ago | JavaScript | Codex · Claude Code |
How does FollowAgents rate this agent?
- Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
- Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
- Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
- Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
- Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
- Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
- No source files were provided for this review; the result is neither a positive nor a negative judgment on this repository.
- Provide revision-pinned README, LICENSE, package., workflows, and representative agent/workflow files to enable a substantive assessment.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
AIWG 2026.10.5 - Session catalog rebuild fixes AIWG 2026.10.5 fixes four problems found while rebuilding session catalogs. Rebuilds now keep history whose source files are gone, discovery uses the providers configured for the workspace, redaction catches credential values written on the line after their key, and aiwg index refresh recovers from a preparation worker that runs out of memory. Rebuild a catalog without losing history A fresh rebuild used to import only sources still on disk, so sessions whose provider files had been pruned disappeared. Pass the previous catalog with …View on GitHub ↗ -
Release notes excerpt
AIWG 2026.10.4 - Inspectable rules and decision-offload visibility AIWG 2026.10.4 makes it easier to inspect the rules that apply to a workspace, see whether bounded decision offload is configured and being used, and preserve that guidance as provider context is regenerated or compacted. It also adds direct signed-tag publication and optional qualification tools, improves deployment performance, and reduces session-import memory use. Inspect the rules for your workspace The scoped rule inventory lists provenance, activation, and fetch paths. You can …View on GitHub ↗ -
v2026.10.3 v2026.10.3 - Direct worker orchestration, safer automation, and security hardening
Release notes excerpt
AIWG 2026.10.3 - Direct worker orchestration, safer automation, and security hardening AIWG 2026.10.3 brings a substantial direct-worker orchestration foundation: model and CLI selection, isolated execution, health reporting, verification, and provider-specific headless profiles. It also improves MCP configuration, unattended CLI workflows, decision visibility, session collection, memory, and deployment diagnostics, alongside a broad CodeQL security sweep. > About the version number: v2026.10.2 was tagged but never published, …View on GitHub ↗