Refunc MCP Server
About
The Refunc MCP Server integrates Refunc serverless functions as tools in the Model Context Protocol (MCP), allowing you to expose functions as MCP endpoints via Server-Sent Events (SSE). It is designed for developers who want to create MCP-compatible tools from existing Refunc…
Details
- Author
- refunc
- Downloads
- 160
- Categories
- Other
Jump to
- Deployable on Kubernetes with a single manifest file.
- Uses refunc-cli to create and manage functions.
- MCP event spec supports multiple tools per function.
- SSE endpoint follows a predictable URL format.
- Tools can be described inline with JSON Schema parameters.
Deploy the Refunc components on Kubernetes using kubectl apply, then proxy the services locally. Install the refunc-cli and initialize a function with rfctl init. Edit the generated lambda.yaml to add an MCP event specification with a token and tool definitions. Finally, update the configuration with rfctl update-config and access the MCP SSE endpoint at the provided URL path.
Refunc MCP Server
Deploy Refunc
Deploy all refunc components in namespace refunc-system.
kubectl apply -f https://raw.githubusercontent.com/refunc/mcp-server/refs/heads/main/deploy/play-all-in-one.yaml
Then proxy refunc service to local.
kubectl port-forward svc/aws-api 8000:http --address 0.0.0.0 -n refunc-system
kubectl port-forward svc/mcp-server 8001:http --address 0.0.0.0 -n refunc-system
PS: The data from the demo deployment is not persistent.
Create Function
Create a Function with refunc-cli.
```
pip install -U refunc-cli
mkdir mcp-demo && cd mcp-demo
rfctl init
Initing function manifest in /path/mcp-demo/lambda.yaml
Name: mcp-demo
Namespace: refunc-system
Select Language:
1 - python
2 - go
Choose from 1, 2 [1]:
Select Runtime:
1 - python3.10
2 - python3.9
3 - python3.8
Choose from 1, 2, 3 [1]:
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.
