> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.recursiv.io/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.recursiv.io/_mcp/server.

# Recursiv vs CrewAI and AutoGPT

## How Does Recursiv Compare to CrewAI and AutoGPT?

CrewAI and AutoGPT are multi-agent frameworks you self-host. You define the agents and roles, then run and operate the whole thing yourself. Recursiv runs the swarm. It gives each agent its own identity, a sandbox, memory and tool approvals, with an audit trail of what ran.

The differentiator is that the multi-agent runtime is the platform. You do not stand up the orchestration, the execution isolation or the governance layer around a self-hosted framework. Recursiv provides them.

## Quick Comparison

| Capability         | Recursiv                  | CrewAI / AutoGPT      |
| ------------------ | ------------------------- | --------------------- |
| Type               | Platform and runtime      | Self-hosted framework |
| Multi-agent        | Swarms and dispatcher     | Crews and agents      |
| Where it runs      | The platform runs it      | Your infrastructure   |
| Per-agent identity | Scoped identity per agent | Build it yourself     |
| Memory             | Brain and memory built in | Bring your own store  |
| Code execution     | Managed sandboxes         | Bring your own        |
| Tool approvals     | Built in                  | Build it yourself     |
| Audit              | Built in                  | Build it yourself     |
| Self-evaluation    | Protocols and simulator   | Build it yourself     |
| Model support      | Any model, swap anytime   | Any model via config  |
| Deploy             | Deployments built in      | Bring your own host   |

## When a Self-Hosted Framework Is the Better Choice

* You want to run the whole multi-agent system inside your own cloud with no platform dependency
* You need full control over the agent loop and crew definitions in your own code
* You are experimenting locally and do not yet need per-agent identity, sandboxes or audit
* Your requirements are met by a framework and you do not expect to need govern or a managed runtime

## Get Started

If you want each agent to have an identity, a sandbox, memory and approvals with an audit trail, start with the [quickstart](/guides/start-here/quickstart).