crewai - 💡(How to fix) Fix [FEATURE] Add Wasm-based code interpreter tool to safely run untrusted code. [2 comments, 2 participants]

Feature Area

Integration with external tools

Is your feature request related to a an existing bug? Please link it here.

It's related to the security vulnerabilities identified in the current CodeInterpreterTool (#4516). The discussion suggests more robust isolation like containers or other non Python-level sandboxing, which is what this integration provides with Wasm.

Describe the solution you'd like

I'm proposing an integration called crewai-capsule. It's a pre-configured version of Capsule, a runtime I built to sandbox AI agent tasks using WebAssembly.

Here's an example of how it would work:

from crewai_capsule import CapsulePythonREPLTool

python_tool = CapsulePythonREPLTool()

code = """
def factorial(n):
    if n <= 1:
        return 1
    return n * factorial(n - 1)

factorial(6)
"""

result = python_tool.run(code)
print(result) # "720"

Only the first run takes about a second (cold start), then every next run starts in ~10ms.

Describe alternatives you've considered

No response

Additional context

PyPI package: crewai-capsule Integration repository: github.com/mavdol/crewAI-capsule Main Capsule repository: github.com/mavdol/capsule

I've also built a separate MCP server that provides the same sandboxing capabilities, if that's of interest: github.com/mavdol/capsule/integrations/mcp-server

I'd be happy to submit a PR to integrate this solution into CrewAI (whether through documentation, a direct integration, or any other convenient way).

Willingness to Contribute

Yes, I'd be happy to submit a pull request