| Filename | Latest commit message | Latest commit date |
|---|---|---|
1a3e5ef459<!-- Provide a brief summary of your changes --> Ensure Go audit-mode interceptors remain non-blocking when validator or mutator invocation returns an error. ## Motivation and Context <!-- Why is this change needed? What problem does it solve? --> SEP says audit-mode interceptors MUST NOT block execution regardless of results. The Go SDK already avoided blocking on audit-mode validation findings, but validator/mutator handler or RPC errors could still abort unless `FailOpen` was set. This change makes audit mode consistently non-blocking for invocation errors as well, matching the existing C# orchestrator behavior where `isAudit || failOpen` allows execution to continue. ## How Has This Been Tested? <!-- Have you tested this in a real application? Which scenarios were tested? --> Added focused Go chain tests covering audit-mode validator and mutator errors. ## Breaking Changes <!-- Will users need to update their code or configurations? --> None. This only relaxes audit-mode error handling to match SEP semantics. ## Types of changes <!-- What types of changes does your code introduce? Put an `x` in all the boxes that apply: --> - [x] Bug fix (non-breaking change which fixes an issue) - [ ] New feature (non-breaking change which adds functionality) - [ ] Breaking change (fix or feature that would cause existing functionality to change) - [ ] Documentation update ## Checklist <!-- Go over all the following points, and put an `x` in all the boxes that apply. --> - [x] I have read the [MCP Documentation](https://modelcontextprotocol.io) - [x] My code follows the repository's style guidelines - [x] New and existing tests pass locally - [x] I have added appropriate error handling - [x] I have added or updated documentation as needed ## Additional context <!-- Add any other context, implementation notes, or design decisions --> C# already treats audit-mode invocation failures as non-blocking. This aligns the Go chain behavior with that implementation and SEP audit-mode semantics. Signed-off-by: Ukjae Jeong <jeongukjae@gmail.com> |
||
| .. | ||
| doc | ||
| examples | ||
| interceptors | ||
| vendor | ||
| go.mod | ||
| go.sum | ||
| README.md | ||
MCP Interceptors - Go Implementation
Go implementation of the MCP Interceptor Extension based on SEP-2624.
Note: Currently the MCP SDK is vendored, in-order to add the Protocol Methods needed for interceptors.
Quick Start
mcpServer := mcp.NewServer(&mcp.Implementation{
Name: "my-server",
Version: "0.1.0",
}, nil)
// Create an extension and register interceptors.
ext := extension.New()
// Register a validator that blocks dangerous tool calls.
ext.AddInterceptor(&interceptors.Validator{
Metadata: interceptors.Metadata{
Name: "block-dangerous",
Hooks: []interceptors.Hook{{
Events: []string{interceptors.EventToolsCall},
Phase: interceptors.PhaseRequest,
}},
Mode: interceptors.ModeEnforce,
},
Handler: func(_ context.Context, inv *interceptors.Invocation) (*interceptors.ValidationResult, error) {
raw := inv.Payload.(json.RawMessage)
var params struct{ Name string `json:"name"` }
json.Unmarshal(raw, ¶ms)
// validate the request...
return &interceptors.ValidationResult{Valid: true}, nil
},
})
// Install on the server and create a chain for middleware.
ext.Install(mcpServer)
chain, err := ext.LocalChain(ctx, mcpServer)
mcpServer.AddReceivingMiddleware(gomiddleware.Middleware(chain))
mcpServer.Run(context.Background(), &mcp.StdioTransport{})
See examples/ for complete working examples.
Documentation
- DESIGN.md — architecture, execution model, integration with the go-sdk.
- PERFORMANCE.md — per-request cost model, allocation summary, and optimization notes.
- CONFORMANCE.md — SEP conformance status.
Package API documentation is available via go doc:
go doc github.com/modelcontextprotocol/ext-interceptors/go/sdk/interceptors