Review · Multi-agent frameworks · August 2026
CrewAI
Fastest way to stage a multi-agent demo. Still thin when grants, budgets, and a scored finish line matter.
6 / 10
Best for · Prototyping role-based crews and showing stakeholders a collaborative agent workflow quickly.
Not for · Teams that need fine-grained runtime control, hard spend caps, or a production agent with isolated permissions on day one.
TL;DR
CrewAI makes multi-agent crews easy to name and hard to own. Role-based agents ship a prototype fast. Production still needs budgets, grants, and a finish line you can score. Hosted Basic is free with a hard 50-execution cap. Enterprise is a sales call. I use it to explore. I do not confuse that with shipping.
Spiked a four-role crew for a research brief; multi-agent tax showed up in token burn.
Product image
CrewAI OG image. Multi-agent mosaics look pretty until a handoff drops a tile.
What it is
The crew.
Python multi-agent framework and hosted AMP platform. You define agents by role and task. Crews orchestrate. Studio adds a visual editor. OSS self-host is free. Hosted path is Basic free tier plus Enterprise custom.
Pricing
What it costs.
| Plan | Price | Note |
|---|---|---|
| CrewAI OSS | $0 | Self-host the framework. LLM usage is on you. |
| Basic (hosted) | Free | Studio, GitHub integration, about 50 workflow executions/month. Hard cap. Confirm on crewai.com/pricing. |
| Enterprise | Custom | SSO, RBAC, VPC options, onboarding. Executions sized to workflow. Mid-tier self-serve Pro was pulled earlier in 2026. |
Prices as of August 2026. Confirm on the vendor page — meters move.
Does well
What works.
Role theater ships fast
Researcher, writer, reviewer. The mental model matches vendor decks. Non-engineers can see a crew move. Useful for buy-in.
Studio and templates
Visual editor, GitHub integration, MCP export on the hosted side. Lower friction than wiring a graph from scratch for a first spike.
Python ecosystem fit
If the team already lives in Python notebooks and wants agent roles without graph theory, CrewAI gets out of the way.
Falls short
What fails.
Multi-agent tax
Four agents often means four times the LLM calls. Handoffs fail quietly. Extra names are not extra intelligence.
Hosted pricing cliff
Free Basic caps around 50 executions. Then Enterprise. The missing middle self-serve tier is a real gap for small teams.
Control surface
Less explicit than LangGraph on durable interrupts and permission isolation. You can add guardrails. You still own production discipline.
Who should
Buy this if.
Builders validating whether a crew shape even helps. Production owners should score the same jobs against one agent with tools before they scale seats.
FAQ
Questions.
- CrewAI or LangGraph?
- CrewAI for fast role-based prototypes. LangGraph when persistence, interrupts, and explicit control flow are the product.
- Is there a cheap paid hosted plan?
- As of August 2026 the public page is Basic free and Enterprise custom. Older Pro tiers were withdrawn. Recheck before you plan a budget.
- Can I run it in production?
- Yes if you add budgets, evals, and tight tools yourself. The framework alone is not a production story.
Also on the desk
Compare next.
Agent frameworks · 8/10
LangGraph
Best default runtime when the agent must pause, resume, and leave an artifact you can score.
Self-hosted automation / agents · 8/10
n8n
The automation runtime technical teams actually self-host: fair-code core, execution-based cloud pricing, and agents that share the same meter.
No-code AI app builder · 6/10
Lovable
Fastest path from prompt to a clickable web app. Credits and hosting meter the real bill. Production money and auth still need an engineer. Score the deployed URL.