Frameworks & Eval · Reviewed 2026-06-16
OpenAcme
STEADY · 61/100
TypeScript multi-agent workforce platform with MCP and multi-provider LLM — solid architecture, but 70 stars means this hasn't been stress-tested outside of the author's use case.
Visit OpenAcme →OpenAcme bills itself as 'an AI workforce' where you define named agents with roles, models, and tools, and the org chart self-delegates from goals to execution. The dependency-aware task scheduler (agents assign work to each other, scheduler wakes dependents when upstream clears) and the mix of web UI and CLI are the differentiating details — most agent frameworks give you either a programmer API or a workflow DSL, not both alongside a visual UI. Multi-provider LLM support (Anthropic, OpenAI, Ollama, OpenRouter) and MCP integration show that the author is building for the current ecosystem, not just a single-vendor solution. TypeScript, pnpm, Node ≥18 is a modern and deployable stack. The concern is scale: 70 GitHub stars is the kind of number that indicates the project was shared in a few communities but hasn't found its audience yet. The question for a framework is always 'does it hold up when 10 other engineers build on top of it' — and at 70 stars, that answer is unverified. The demo GIF in the README is a genuine positive (shows real operation, not just a feature list).
Why STEADY
STEADY (61) because the architecture is sound (org-chart delegation, dependency scheduler, multi-provider, MCP), the demo is real, and the distribution (npm @openacme/cli, web + CLI) shows product thinking. Not higher because 70 stars = very early adoption and production reliability is unverified. Borderline between STEADY and FADING — held STEADY because the architecture quality is above average for the star count.
What it does well
- Org-chart delegation model — hierarchy-aware task assignment, not just round-robin agents
- Task dependency scheduler — agent B wakes when agent A completes, not just polling
- Multi-provider LLM: Anthropic, OpenAI, Ollama, OpenRouter — no vendor lock-in
- Web UI + CLI in the same product — accessible to both visual and CLI users
- MCP integration out of the box — tools can be MCP-served endpoints
What it fails at
- 70 GitHub stars = very early; production reliability at scale is unverified
- Hosted option requires building from source — no pre-built binary or hosted SaaS
- Memory/context management across agents not documented on public surface
- CI status in README exists but GitHub login wall blocked verification
- 70 stars may indicate the delegation model is more complex to configure than the README implies
Best for
- TypeScript developers wanting a multi-agent orchestration framework with visual oversight
- Projects with clear hierarchical task structures where delegation is natural
- Teams wanting multi-provider LLM flexibility without rewriting agent definitions
- Developers who want to evaluate org-chart-style agent delegation before committing to a more established framework
Not recommended for
- Teams needing production-proven stability — 70 stars isn't enough for enterprise adoption without internal testing
- Non-TypeScript teams
- Projects with flat task structures where org-chart delegation adds overhead without value
- Organizations that need commercial support or SLAs
Compared to
-
mybono-ai-orchestrator
general-workforce-platform
Both are TypeScript multi-agent frameworks at early stage. ai-orchestrator focuses specifically on Claude Code + Ollama cost optimization. OpenAcme is a more general multi-provider workforce platform with a web UI. OpenAcme for visual oversight and multi-provider support; ai-orchestrator for Claude Code cost optimization.
Agent relevance
API CLI MCP SDK
TypeScript SDK + CLI. Define agents, tools, and roles programmatically. MCP-served tools plug in directly. Web UI provides visual oversight. Other agents can interact via the task API. Node ≥18 required.
Agent-friendly score: 8/10
Evidence
Public-surface checklist
- ✓ homepage_loads (required)
- ✓ primary_value_prop (required) — 'An AI workforce. You're in charge.'
- ✓ cta_present (required) — GitHub link + npm install instructions
- ✓ pricing_or_access — Free and open source (MIT)
- ✓ evidence_or_demo — Demo GIF in README showing real agent delegation