Agency Swarm
Agent agencies with strict communication — 4.6K stars, MIT, actively maintained.
About
Agency Swarm (4,598 GitHub stars, MIT) is a multi-agent framework organized as agencies: agents with defined roles communicate through a strict agency chart, with a visual builder for designing the organization.
Actively maintained (pushed Oct 2026). The agency metaphor (CEO agents delegating to specialist agents) suits business-process automation. Smaller than the majors, but the visual agency builder is distinctive. For teams that think in org charts, this maps naturally.
What it can do
- • Agency-based multi-agent organization\n• Strict inter-agent communication\n• Visual agency builder\n• Role-based delegation\n• Tool integration per agent
Setup
- Install:
pip install agency-swarm.\n2. Set your model API key.\n3. Define agents and an agency chart; run the agency.\nSee the GitHub README for the visual builder.
Prerequisites
Python ≥ 3.10; an LLM API key (OpenAI recommended).
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier). MIT, actively maintained with 27 releases and only 6 open issues, with a distinctive visual agency builder. Docked two points: (1) the agency metaphor is niche; (2) OpenAI-centric. For org-chart-style agent design.
★ 4,598 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 9.9/10
- Openness · license + docs
- 9.1/10
- Manual review · sanity read + red-flag check
- 6.0/10
- · 4,598 GitHub stars (log-scaled vs catalog peers)
- · last push 1d ago; 27 releases/12mo; 6 open issues
- · manual review: distinctive visual builder; niche metaphor; OpenAI-centric; limited production tooling; license: MIT
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.