### Context
Design (`design/mrtr.md`) and implementation proposal for Multi
Round-Trip Requests (MRTR) per
[SEP-2322](de6d76fba3/seps/2322-MRTR.md).
### Changes
- Introduce `InputRequest` and `InputResponse` sealed interfaces with
corresponding `InputRequestMap` and `InputResponseMap` types for custom
map JSON codec implementation.
- Extend `CallToolParams`, `GetPromptParams`, and `ReadResourceParams`
with `InputResponses`, `RequestState` and an unexported `resultType`
("complete" or "input_required") for retry round-trips. The type is set
by the SDK based on input requests presence.
- Add client-side MRTR middleware enabled by default that automatically
fulfills input requests and retries.
- Added MRTROptions on ClientOptions to allow disabling the middleware
and configuring the number of retries.
- Add server-side MRTR middleware for backward compatibility:
transparently bridges input requests to direct
Elicit/CreateMessage/ListRoots calls for older clients
Since ping is optional per the MCP spec, a peer that doesn't implement
it will respond with a -32601 (method not found) error. Previously,
startKeepalive treated all ping errors identically, closing the session
even when the peer was perfectly healthy.
Now, a method-not-found response stops the keepalive goroutine without
closing the session.
Fixes#897
## Summary
Two error sites in the MCP runtime were previously unobservable:
- **`mcp/shared.go` `startKeepalive`** — when a keepalive ping failed,
the session was closed silently with no indication of why.
- **`mcp/transport.go` `connect`** — `jsonrpc2.OnInternalError` printed
via the package-level `log.Printf`, bypassing the `*slog.Logger`
configured on `ServerOptions`/`ClientOptions`.
Both sites now report through the existing `*slog.Logger` that `Server`
and `Client` already guarantee non-nil via `ensureLogger`. The logger is
threaded into the unexported `startKeepalive` and `connect` helpers as a
parameter — **no new public API surface**.
## Context
Per @jba's [suggestion on
#865](https://github.com/modelcontextprotocol/go-sdk/pull/865#issuecomment-4156166474):
> Maybe we can just use the logger in all the places we currently drop
the error. Then it's still reported, but we don't have to add any API.
This PR is that smaller alternative. It addresses the concrete
observability gap in #218 without expanding `ServerOptions`. PR #865
(the `ErrorHandler` callback proposal) remains open pending community
input on whether a structured callback API is also wanted on top of
logging.
`notifySessions` already logs at `Warn` level on delivery errors and is
intentionally left untouched.
## Test plan
- [x] New regression test `TestKeepAliveFailure_Logged` asserts the
keepalive failure produces a log line on the configured logger.
- [x] Verified the test fails when the new `logger.Error(...)` call is
removed (sanity check that it pins the behavior).
- [x] `go test ./mcp/... -count=1` passes.
- [x] `go vet ./...` clean.
- [x] `go build ./...` clean.
Refs #218
---------
Co-authored-by: Maciej Kisiel <mkisiel@google.com>
SetError now only populates the Content field when it has not already
been populated. This allows callers to provide a user-friendly message
while still recording the underlying error for middleware inspection
via GetError.
The len check (rather than nil check) ensures both nil and empty-slice
Content are treated as "not yet populated."
Fixes#858
Converted timing-based tests to use Go 1.25's synctest for instant,
deterministic execution with simulated time.
Moved converted tests to *_go125_test.go files. Tests using real I/O
(exec.Command, httptest) remain timing-based as they require actual
system operations.
Fixes#680
---------
Co-authored-by: i586147 <lakshana.goswami@sap.com>
Debounce and batch notifications for changes in server feature sets
(adding or removing tools, prompts, resources, or resource templates).
This PR handles server-side change notifications only. We could do the
same for client-side notifications (e.g. roots changed), but at the
moment we don't need to.
Fixes#649.
This adds a simple schema and checks that the content returned by an
elicitation handler is validated against the schema. Previously, we had
no test coverage checking if the content was validated (i.e., we would
see no test failures if we removed content validation against the
schema).
The bulk of the actual validation logic is handled by the jsonschema
library so adding comprehensive tests covering different schema-features
is redundant, we simply need to see that some validation is happening.
See https://github.com/modelcontextprotocol/go-sdk/issues/625 for more
details on the motivation of this change.
Fixes#625
---------
Co-authored-by: Markus Kusano <kusano@google.com>
Expose the jsonrpc.Error type to allow access to underlying JSON-RPC
error codes. Also, expose common JSON-RPC error codes in the jsonrpc
package, and MCP error codes in the mcp package.
Also, implement support for the MCP URL elicitation required error
(-32042), enabling servers to request out-of-band user authorization and
clients to automatically handle these requests with retry logic.
Server-side:
- Add CodeURLElicitationRequired constant
- Add URLElicitationRequiredError() constructor.
Client-side:
- Add urlElicitationMiddleware to intercept URLElicitationRequired
errors and retry calls.
- Update callElicitationCompleteHandler to signal waiting operations.
- Fix capability advertisement to support both form and URL modes.
Fixes#452Fixes#623
Add initial support for URL mode elicitation, by introducing the new
types, as well as both server and client-side validation.
A follow-up CL will add support for ErrURLElicitationRequired.
For #623
This PR fixes goroutine leaks in all unit tests.
To find the leaks I integrated https://github.com/uber-go/goleak and I
suggest to keep using it to catch any future regressions.
I used the folowing script to more easily find individual tests which
were leaking goroutines:
```
for t in $(go test -run=Nothing -list=. ./mcp | grep -v ok); do go test ./mcp -run="$t" > /dev/null && echo -n . || echo -e "\nDetected leak: $t"; done
```
Provide a way for middleware to get the error from a tool call,
demonstrating that we can add this functionality without making a
breaking change.
Leave getError unexported for now; we can export it (and setError) at
any time.
For #64.
Add a framework for feature documentation, and start populating it with
our SDK documentation.
This framework is as follows:
- internal/docs/**.src.md is the markdown source for the docs/
directory.
- The x/example/internal/cmd/weave tool is used to compile these docs to
the top-level docs, supporting both linked code samples and generated
tables of contents.
- The readme-check workflow is updated to check these docs as well.
- The structure of these docs follows the MCP spec.
- Wherever possible, example code is linked from actual Go documentation
examples, so that it is testable.
Some minor modifications to the weave tool were made to support this
framework.
Additionally, partially fill out this documentation with content on
base protocol and client features, as well as troubleshooting help.
Along the way, a bug was encountered that our LoggingTransport was not
concurrency safe. This is fixed with a mutex.
Fixes https://github.com/modelcontextprotocol/go-sdk/issues/466
Fixes https://github.com/modelcontextprotocol/go-sdk/issues/409
Updates https://github.com/modelcontextprotocol/go-sdk/issues/442
Our validation logic was avoiding double-unmarshalling as much as
possible, by parsing before validation and validating the Go type.
This only works if the Go type has the same structure as its JSON
representation, which may not be the case in the presence of types with
custom MarshalJSON or UnmarshalJSON methods (such as time.Time).
But even if the Go type doesn't use any custom marshalling, validation
is broken, because we can't differentiate zero values from missing
values.
Bite the bullet and use double-unmarshalling for both input and output
schemas. Coincidentally, this fixes three bugs:
- We were accepting case-insensitive JSON keys, since we parsed first,
even though they should have been rejected. A number of tests were
wrong.
- Defaults were overriding present-yet-zero values, as noted in an
incorrect test case.
- When "arguments" was missing, validation wasn't performed, no defaults
were applied, and unmarshalling failed even if all properties were
optional.
First unmarshalling to map[string]any allows us to fix all these bugs.
Unfortunately, it means a 3x increase in the number of reflection
operations (we need to unmarshal, apply defaults and validate,
re-marshal with the defaults, and then unmarshal into the Go type).
However, this is not likely to be a significant overhead, and we can
always optimize in the future.
Update github.com/google/jsonschema-go to pick up necessary improvements
supporting this change.
Additionally, fix the error codes for invalid tool parameters, to be
consistent with other SDKs (Invalid Params: -32602).
Fixes#447Fixes#449
Updates #450
Upon a closer reading of the spec, it is OK for the server to response
to client requests before it has received notifications/initialized (as
long as it has received initialize). This effectively rolls back the fix
from #225.
Some hooks are left for enforcing strictness around server->client
requests prior to initialized. These will be revisited in subsequent
CLs.
For #395
The spec makes it clear that input schemas must have type "object".
Enforce that.
Allow "any" as an input argument type by special-casing it.
Fixes#283.
Replace all occurrences of ServerRequest[*CallToolParams] and other
concrete instantiations with CallToolRequest and the like.
Make the XXXRequest types aliases, to preserve the convenience of
generics for the internal machinery (see shared.go, for example.)
I will expand the aliases in a followup PR.
As reported in #199 and #200, the fact that we return a possibly "null"
schema for pointer types breaks various clients, which expect schemas to
be of type "object". This is an unfortunate footgun. For now, assume
that the user wants us to treat pointers equivalently to non-pointers.
If we want to change this behavior in the future, we can do so behind an
option.
+ a test
Also fix the handling of nil results in the case where the output schema
is non-nil: we must provide structured content in this case. (This was
causing the test to fail).
Fixes#199Fixes#200
Change the design of tool handlers by removing genericity from the common path.
- `ToolHandler` gets the args as a `json.RawMessage`. Using Server.AddTool does no unmarshaling or schema validation.
- `ToolHandlerFor`, installed with `AddTool`, are the only generic pieces.
TODO:
uncomment and fix some tests
Implement the concurrency model described in #26: notifications are
synchronous, but calls are asynchronous (except for 'initialize'). To
achieve this, implement jsonrpc2.Async(ctx) to signal asynchronous
handling. This is simpler to use than returning ErrAsyncResponse and
calling Respond, and since this is an internal detail we don't need to
worry too much about whether it's idiomatic.
Add tests that verify both features, for both client and server.
Also:
- replace req.ID.IsValid with req.IsCall
- remove the methodHandler type as we can just use MethodHandler
Fixes#26
In #202, I added the checkRequest helper to validate incoming requests,
and invoked it in the stremable transports to preemptively reject
invalid HTTP requests, so that a jsonrpc error could be translated to an
HTTP error.
However, this introduced a bug: since cancellation was handled in the
jsonrpc2 layer, we never had to validate it in the mcp layer, and
therefore never added methodInfo. As a result, it was reported as an
invalid request in the http layer.
Add a test, and a fix. The simplest fix was to create stubs that are
placeholders for cancellation.
This was discovered in the course of investigating #285.
Handlers and client/server methods take a single request argument
combining the session and params. This slightly simplifies the signature
for handlers, since there is no longer an often-ignored session
argument.
But more importantly, it opens the door to adding more information in
requests, such as auth info and HTTP request headers.
For #243.
As described in #272, there is really no reason for transports to be
closed structs with constructors, since their state must be established
by a call to Connect. Making them open structs simplifies their APIs,
and means that all transports can be extended in the future: we don't
have to create empty Options structs just for the purpose of future
compatibility.
For now, the related constructors and options structs are simply
deprecated (with go:fix directives where possible). A future CL will
remove them prior to the v1.0.0 release.
For #272
Support stateless streamable sessions by adding a GetSessionID function
to StreamableHTTPOptions. If GetSessionID returns "", the session is
stateless, and no validation is performed. This is implemented by
providing the session a trivial initialization state.
To implement this, some parts of #232 (distributed sessions) are copied
over, since they add an API for creating an already-initialized session.
In total, the following new API is added:
- StreamableHTTPOptions.GetSessionID
- ServerSessionOptions (a new parameter to Server.Connect)
- ServerSessionState
- ClientSessionOptions (a new parameter to Client.Connect, for symmetry)
For #10
## Background
Since the Server has a map of ServersSessions->subscriptions, and takes
care of routing ResourceUpdate messages to ServerSessions, I think the
decision was made to leave *ServerSession out of the
SubscribeHandler/UnsubscribeHandler signatures. However, resource
subscriptions are not specific to the a particular client and this
information is kept private in the Session today.
## Proposal
Align the Subscribe/Unsubscribe handlers with the other two handlers
that are client specific (ProgressNotificationHandler, and
RootListChangeHanlder), and include *ServerSession in the signature.
It comes in practice because because a gateway server that needs to
forward subscriptions still needs this information. The routing logic in
Server is convenient and means that most servers won't need to know
which ServerSessions are subscribed. But I think there are still use
cases where the user will need to know.
When calling `ServerSession.CreateMessage`, will get an json unmarshal
error because `CreateMessageResult.Content` is an interface, and it
doesn't implement `json.Unmarshaller`.
Implement the `json.Unmarshaller` to `CreateMessageResult`, let it able
to unmarshal the client response without error
BREAKING CHANGE
Change the API for adding tools, prompts, resources and resource
templates to a server.
The ServerXXX types are gone along with the plural Add methods
(AddTools, AddPrompts etc.)
Instead there are single Add methods (AddTool, AddPrompt).
In addition, instead of NewServerTool, there is AddTool[In, Out].
Fixes#73.
