AC Framework
AC Framework enables AI assistants to program better with advanced methodologies and Spec-Driven Development (SDD). Structured workflows for superior code quality.
Built-in SDD workflow that guides AI through structured development phases for higher quality code.
Enforces best practices, design patterns, and architectural principles for maintainable code.
acfm init configures your project with SDD workflows and methodology rules instantly.
Built-in verification ensures AI-generated code meets specs with testing and validation workflows.
Keep maximum quality with two execution profiles: high-load optimization for complex changes, and lite mode for lower token usage with strict quality gates.
Full-depth execution for large codebases, complex refactors, and high-risk architectural changes.
Loads only essential workflow skills first, then activates additional skills only when risk gates are triggered.
AC Framework keeps complete compatibility with OpenSpec workflows while using the built-in acfm spec commands.
.acfm/ and legacy openspec/ directories.AC Framework combines structured implementation workflows, collaborative multi-agent execution, and persistent memory with MCP integrations so your assistants behave like a coordinated engineering team.
SynapseGrid runs a real role-based engineering loop with planner, critic, coder, and reviewer in coordinated rounds, where each role sees accumulated context.
acfm agents start --task "Implement secure auth" --mux auto
Every session is persisted with readable logs, summaries, diagnostics, and raw captures, so users can inspect what each agent said and did.
acfm agents artifacts --watch --interval 1200
NexusVault gives your assistants long-lived memory so architecture decisions, bugfix patterns, conventions, and context survive across sessions.
~/.acfm/memory.db.acfm memory init
acfm memory recall
acfm memory search "authentication refresh token"
AC Framework exposes both memory and collaborative runtimes through MCP servers, while keeping implementation disciplined with spec-driven workflows.
ac-framework-memory and ac-framework-collab..acfm/ and legacy openspec/.acfm spec init
acfm spec new add-rate-limiting
acfm agents model choose
acfm agents runtime install-zellij
acfm agents runtime set zellij
acfm agents doctor
acfm agents transcript --role all --limit 40
acfm agents summary
acfm agents export --format md --out session.md
Security and reliability by design: artifact creation is runtime-owned (not prompt-dependent), external links are isolated with safe target attributes, and session observability is deterministic across running, failed, and stopped states.
Works with your favorite AI coding tools
Get started in seconds with a single command
Requires Node.js >= 18.0.0