modelcontextprotocol/docs
Watch
1
0
Fork
You've already forked docs
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge pull request #18 from modelcontextprotocol/justin/concepts

Browse source 
Revise "Concepts" section, stub tabs to switch to Python
This commit is contained in:
Justin Spahr-Summers 2024-11-21 15:52:48 +00:00 • committed by GitHub
commit 1363e2c60a
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 681 additions and 706 deletions
Download patch file Download diff file Expand all files Collapse all files

242
docs/concepts/architecture.mdx
View file

@ -3,49 +3,69 @@ title: "Core architecture"
description: "Understand how MCP connects clients, servers, and LLMs"
---
The Model Context Protocol (MCP) is built on a flexible, extensible architecture that enables seamless communication between LLM applications and external services. This document covers the core architectural components and concepts.
The Model Context Protocol (MCP) is built on a flexible, extensible architecture that enables seamless communication between LLM applications and integrations. This document covers the core architectural components and concepts.
## Overview
MCP follows a client-server architecture where:
- **Clients** are LLM applications (like claude.ai or IDEs) that initiate connections
- **Hosts** are LLM applications (like Claude Desktop or IDEs) that initiate connections
- **Clients** maintain 1:1 connections with servers, inside the host application
- **Servers** provide context, tools, and prompts to clients
- **Hosts** are processes that run MCP clients
```mermaid
flowchart LR
subgraph "Host (e.g., claude.ai)"
client[MCP Client]
subgraph " Host (e.g., Claude Desktop) "
client1[MCP Client]
client2[MCP Client]
end
subgraph "Server Process"
server[MCP Server]
server1[MCP Server]
end
client <-->|Transport Layer| server
subgraph "Server Process"
server2[MCP Server]
end
client1 <-->|Transport Layer| server1
client2 <-->|Transport Layer| server2
```
## Core components
### Protocol layer
The protocol layer handles message framing, request/response linking, and high-level communication patterns. Key classes include:
The protocol layer handles message framing, request/response linking, and high-level communication patterns.
```typescript
class Protocol<SendRequestT, SendNotificationT, SendResultT> {
// Handle incoming requests
setRequestHandler<T>(schema: T, handler: (request: T) => Result): void
// Handle incoming notifications
setNotificationHandler<T>(schema: T, handler: (notification: T) => void): void
// Send requests and await responses
request<T>(request: Request, schema: T): Promise<T>
// Send one-way notifications
notification(notification: Notification): Promise<void>
}
```
<Tabs>
<Tab title="TypeScript">
```typescript
class Protocol<Request, Notification, Result> {
// Handle incoming requests
setRequestHandler<T>(schema: T, handler: (request: T, extra: RequestHandlerExtra) => Promise<Result>): void
// Handle incoming notifications
setNotificationHandler<T>(schema: T, handler: (notification: T) => Promise<void>): void
// Send requests and await responses
request<T>(request: Request, schema: T, options?: RequestOptions): Promise<T>
// Send one-way notifications
notification(notification: Notification): Promise<void>
}
```
</Tab>
<Tab title="Python">
```python
# Python implementation coming soon
```
</Tab>
</Tabs>
Key classes include:
* `Protocol`
* `Client`
* `Server`
### Transport layer
@ -54,66 +74,48 @@ The transport layer handles the actual communication between clients and servers
1. **Stdio transport**
- Uses standard input/output for communication
- Ideal for local processes
```typescript
class StdioTransport implements Transport {
constructor(stdin: Readable, stdout: Writable)
send(message: JSONRPCMessage): Promise<void>
}
```
2. **WebSocket transport**
- Enables real-time bidirectional communication
- Supports remote connections
```typescript
class WebSocketTransport implements Transport {
constructor(url: URL)
send(message: JSONRPCMessage): Promise<void>
}
```
3. **SSE transport**
2. **HTTP with SSE transport**
- Uses Server-Sent Events for server-to-client messages
- HTTP POST for client-to-server messages
```typescript
class SSETransport implements Transport {
constructor(url: URL)
send(message: JSONRPCMessage): Promise<void>
}
```
All transports use [JSON-RPC](https://www.jsonrpc.org/) 2.0 to exchange messages. See the [specification](https://spec.modelcontextprotocol.io) for detailed information about the Model Context Protocol message format.
### Message types
MCP uses JSON-RPC 2.0 as its message format with three main types:
MCP has these main types of messages:
1. **Requests**
```typescript
interface Request {
method: string
params?: {
_meta?: {
progressToken?: string | number
}
}
}
```
1. **Requests** expect a response from the other side:
```typescript
interface Request {
method: string;
params?: { ... };
}
```
2. **Notifications**
```typescript
interface Notification {
method: string
params?: {
_meta?: Record<string, unknown>
}
}
```
2. **Notifications** are one-way messages that don't expect a response:
```typescript
interface Notification {
method: string;
params?: { ... };
}
```
3. **Results**
```typescript
interface Result {
_meta?: Record<string, unknown>
[key: string]: unknown
}
```
3. **Results** are successful responses to requests:
```typescript
interface Result {
[key: string]: unknown;
}
```
4. **Errors** indicate that a request failed:
```typescript
interface Error {
code: number;
message: string;
data?: unknown;
}
```
## Connection lifecycle
@ -123,27 +125,25 @@ MCP uses JSON-RPC 2.0 as its message format with three main types:
sequenceDiagram
participant Client
participant Server
Client->>Server: initialize request
Server->>Client: initialize response
Client->>Server: initialized notification
Note over Client,Server: Connection ready for use
```
1. Client sends `initialize` request with capabilities
2. Server responds with supported features
3. Client sends `initialized` notification
1. Client sends `initialize` request with protocol version and capabilities
2. Server responds with its protocol version and capabilities
3. Client sends `initialized` notification as acknowledgment
4. Normal message exchange begins
### 2. Message exchange
After initialization, the following patterns are supported:
- **Request-Response**: Client sends request, server responds
- **Request-Response**: Client or server sends requests, the other responds
- **Notifications**: Either party sends one-way messages
- **Progress Updates**: Long-running operations can report progress
- **Resource Updates**: Servers notify about changed resources
### 3. Termination
@ -154,19 +154,21 @@ Either party can terminate the connection:
## Error handling
MCP defines standard error codes and handling:
MCP defines these standard error codes:
```typescript
enum ErrorCode {
ConnectionClosed = -1,
ParseError = -32700,
InvalidRequest = -32600,
MethodNotFound = -32601,
InvalidParams = -32602,
InternalError = -32603
// Standard JSON-RPC error codes
ParseError = -32700,
InvalidRequest = -32600,
MethodNotFound = -32601,
InvalidParams = -32602,
InternalError = -32603
}
```
SDKs and applications can define their own error codes above -32000.
Errors are propagated through:
- Error responses to requests
- Error events on transports
@ -176,31 +178,44 @@ Errors are propagated through:
Here's a basic example of implementing an MCP server:
```typescript
import { Server } from "@modelcontextprotocol/sdk/server";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio";
<Tabs>
<Tab title="TypeScript">
```typescript
import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
const server = new Server({
name: "example-server",
version: "1.0.0"
});
const server = new Server({
name: "example-server",
version: "1.0.0"
}, {
capabilities: {
resources: {}
}
});
// Handle requests
server.setRequestHandler(ListResourcesRequestSchema, async () => {
return {
// Handle requests
server.setRequestHandler(ListResourcesRequestSchema, async () => {
return {
resources: [
{
uri: "example://resource",
name: "Example Resource"
}
{
uri: "example://resource",
name: "Example Resource"
}
]
};
});
};
});
// Connect transport
const transport = new StdioServerTransport();
await server.connect(transport);
```
// Connect transport
const transport = new StdioServerTransport();
await server.connect(transport);
```
</Tab>
<Tab title="Python">
```python
# Python implementation coming soon
```
</Tab>
</Tabs>
## Best practices
@ -212,9 +227,8 @@ await server.connect(transport);
- Simple process management
2. **Remote communication**
- WebSocket for full-duplex communication
- SSE for scenarios requiring HTTP compatibility
- Consider security implications
- Use SSE for scenarios requiring HTTP compatibility
- Consider security implications including authentication and authorization
### Message handling
@ -277,4 +291,4 @@ await server.connect(transport);
- Test different transports
- Verify error handling
- Check edge cases
- Load test servers
- Load test servers

366
docs/concepts/prompts.mdx
View file

@ -3,7 +3,11 @@ title: "Prompts"
description: "Create resuable prompt templates and workflows"
---
MCP Prompts enable servers to define reusable prompt templates and workflows that clients can easily surface to users and LLMs. They provide a powerful way to standardize and share common LLM interactions.
Prompts enable servers to define reusable prompt templates and workflows that clients can easily surface to users and LLMs. They provide a powerful way to standardize and share common LLM interactions.
<Note>
Prompts are designed to be **user-controlled**, meaning they are exposed from servers to clients with the intention of the user being able to explicitly select them for use.
</Note>
## Overview
@ -20,13 +24,13 @@ Each prompt is defined with:
```typescript
{
name: string; // Unique identifier for the prompt
description?: string; // Human-readable description
arguments?: [ // Optional list of arguments
name: string; // Unique identifier for the prompt
description?: string; // Human-readable description
arguments?: [ // Optional list of arguments
{
name: string; // Argument identifier
description?: string; // Argument description
required?: boolean; // Whether argument is required
name: string; // Argument identifier
description?: string; // Argument description
required?: boolean; // Whether argument is required
}
]
}
@ -78,15 +82,13 @@ To use a prompt, clients make a `prompts/get` request:
// Response
{
description?: string, // Optional context about the prompt
messages: [ // Messages to send to the LLM
description: "Analyze Python code for potential improvements",
messages: [
{
role: "user" | "assistant",
role: "user",
content: {
type: "text" | "image",
text?: string,
data?: string, // base64 for images
mimeType?: string
type: "text",
text: "Please analyze the following Python code for potential improvements:\n\n```python\ndef calculate_sum(numbers):\n total = 0\n for num in numbers:\n total = total + num\n return total\n\nresult = calculate_sum([1, 2, 3, 4, 5])\nprint(result)\n```"
}
}
]
@ -97,57 +99,63 @@ To use a prompt, clients make a `prompts/get` request:
Prompts can be dynamic and include:
### Context integration
### Embedded resource context
```typescript
const analyzeLogsPrompt = {
name: "analyze-logs",
description: "Analyze recent system logs",
arguments: [
```json
{
"name": "analyze-project",
"description": "Analyze project logs and code",
"arguments": [
{
name: "timeframe",
description: "Time period to analyze",
required: true
"name": "timeframe",
"description": "Time period to analyze logs",
"required": true
},
{
"name": "fileUri",
"description": "URI of code file to review",
"required": true
}
]
};
server.setRequestHandler(GetPromptRequestSchema, async (request) => {
if (request.params.name === "analyze-logs") {
const logs = await getSystemLogs(request.params.arguments?.timeframe);
return {
messages: [
{
role: "user",
content: {
type: "text",
text: `Analyze these system logs and identify any issues:\n\n${logs}`
}
}
]
};
}
});
}
```
### Resource references
When handling the `prompts/get` request:
```typescript
const codeReviewPrompt = {
name: "review-code",
async getMessages(fileUri: string) {
const code = await readResource(fileUri);
return [
{
role: "user",
content: {
type: "text",
text: `Review this code for best practices:\n\n${code}`
```json
{
"messages": [
{
"role": "user",
"content": {
"type": "text",
"text": "Analyze these system logs and the code file for any issues:"
}
},
{
"role": "user",
"content": {
"type": "resource",
"resource": {
"uri": "logs://recent?timeframe=1h",
"text": "[2024-03-14 15:32:11] ERROR: Connection timeout in network.py:127\n[2024-03-14 15:32:15] WARN: Retrying connection (attempt 2/3)\n[2024-03-14 15:32:20] ERROR: Max retries exceeded",
"mimeType": "text/plain"
}
}
];
}
};
},
{
"role": "user",
"content": {
"type": "resource",
"resource": {
"uri": "file:///path/to/code.py",
"text": "def connect_to_service(timeout=30):\n retries = 3\n for attempt in range(retries):\n try:\n return establish_connection(timeout)\n except TimeoutError:\n if attempt == retries - 1:\n raise\n time.sleep(5)\n\ndef establish_connection(timeout):\n # Connection implementation\n pass",
"mimeType": "text/x-python"
}
}
}
]
}
```
### Multi-step workflows
@ -187,94 +195,107 @@ const debugWorkflow = {
Here's a complete example of implementing prompts in an MCP server:
```typescript
import { Server } from "@modelcontextprotocol/sdk/server";
import {
ListPromptsRequestSchema,
GetPromptRequestSchema
} from "@modelcontextprotocol/sdk/types";
<Tabs>
<Tab title="TypeScript">
```typescript
import { Server } from "@modelcontextprotocol/sdk/server";
import {
ListPromptsRequestSchema,
GetPromptRequestSchema
} from "@modelcontextprotocol/sdk/types";
const PROMPTS = {
"git-commit": {
name: "git-commit",
description: "Generate a Git commit message",
arguments: [
{
name: "changes",
description: "Git diff or description of changes",
required: true
}
]
},
"explain-code": {
name: "explain-code",
description: "Explain how code works",
arguments: [
{
name: "code",
description: "Code to explain",
required: true
const PROMPTS = {
"git-commit": {
name: "git-commit",
description: "Generate a Git commit message",
arguments: [
{
name: "changes",
description: "Git diff or description of changes",
required: true
}
]
},
{
name: "language",
description: "Programming language",
required: false
"explain-code": {
name: "explain-code",
description: "Explain how code works",
arguments: [
{
name: "code",
description: "Code to explain",
required: true
},
{
name: "language",
description: "Programming language",
required: false
}
]
}
]
}
};
const server = new Server({
name: "example-prompts-server",
version: "1.0.0"
});
// List available prompts
server.setRequestHandler(ListPromptsRequestSchema, async () => {
return {
prompts: Object.values(PROMPTS)
};
});
// Get specific prompt
server.setRequestHandler(GetPromptRequestSchema, async (request) => {
const prompt = PROMPTS[request.params.name];
if (!prompt) {
throw new Error(`Prompt not found: ${request.params.name}`);
}
if (request.params.name === "git-commit") {
return {
messages: [
{
role: "user",
content: {
type: "text",
text: `Generate a concise but descriptive commit message for these changes:\n\n${request.params.arguments?.changes}`
}
}
]
};
}
if (request.params.name === "explain-code") {
const language = request.params.arguments?.language || "Unknown";
return {
messages: [
{
role: "user",
content: {
type: "text",
text: `Explain how this ${language} code works:\n\n${request.params.arguments?.code}`
}
}
]
};
}
throw new Error("Prompt implementation not found");
});
```
const server = new Server({
name: "example-prompts-server",
version: "1.0.0"
}, {
capabilities: {
prompts: {}
}
});
// List available prompts
server.setRequestHandler(ListPromptsRequestSchema, async () => {
return {
prompts: Object.values(PROMPTS)
};
});
// Get specific prompt
server.setRequestHandler(GetPromptRequestSchema, async (request) => {
const prompt = PROMPTS[request.params.name];
if (!prompt) {
throw new Error(`Prompt not found: ${request.params.name}`);
}
if (request.params.name === "git-commit") {
return {
messages: [
{
role: "user",
content: {
type: "text",
text: `Generate a concise but descriptive commit message for these changes:\n\n${request.params.arguments?.changes}`
}
}
]
};
}
if (request.params.name === "explain-code") {
const language = request.params.arguments?.language || "Unknown";
return {
messages: [
{
role: "user",
content: {
type: "text",
text: `Explain how this ${language} code works:\n\n${request.params.arguments?.code}`
}
}
]
};
}
throw new Error("Prompt implementation not found");
});
```
</Tab>
<Tab title="Python">
```python
# Python implementation coming soon
```
</Tab>
</Tabs>
## Best practices
@ -324,70 +345,3 @@ When implementing prompts:
- Implement timeouts
- Consider prompt injection risks
- Document security requirements
## Advanced features
### Prompt chaining
```typescript
const analyzePRPrompt = {
async getMessages(prId: string) {
// Chain multiple prompts
const diffPrompt = await getPrompt("analyze-diff");
const testsPrompt = await getPrompt("review-tests");
const securityPrompt = await getPrompt("security-review");
return [
...diffPrompt.messages,
...testsPrompt.messages,
...securityPrompt.messages
];
}
};
```
### Context enrichment
```typescript
const troubleshootPrompt = {
async getMessages(error: string) {
// Enrich with system context
const systemInfo = await getSystemInfo();
const recentLogs = await getRecentLogs();
const stackTrace = await parseError(error);
return [
{
role: "user",
content: {
type: "text",
text: `Help troubleshoot this error:
Error: ${error}
System: ${systemInfo}
Logs: ${recentLogs}
Stack: ${stackTrace}`
}
}
];
}
};
```
### Interactive workflows
```typescript
const debugWorkflow = {
async getMessages(stage: string, context: any) {
switch (stage) {
case "initial":
return promptForError();
case "analysis":
return analyzeError(context.error);
case "solution":
return suggestFixes(context.analysis);
case "verification":
return verifyFix(context.solution);
}
}
};
```

125
docs/concepts/resources.mdx
View file

@ -5,6 +5,12 @@ description: "Expose data and content from your servers to LLMs"
Resources are a core primitive in the Model Context Protocol (MCP) that allow servers to expose data and content that can be read by clients and used as context for LLM interactions.
<Note>
Resources are designed to be **application-controlled**, meaning that the client application can decide how and when they should be used.
For example, one application may require users to explicitly select resources, while another could automatically select them based on heuristics or even at the discretion of the AI model itself.
</Note>
## Overview
Resources represent any kind of data that an MCP server wants to make available to clients. This can include:
@ -24,7 +30,7 @@ Each resource is identified by a unique URI and can contain either text or binar
Resources are identified using URIs that follow this format:
```
[protocol]://[path]
[protocol]://[host]/[path]
```
For example:
@ -42,7 +48,7 @@ Resources can contain two types of content:
Text resources contain UTF-8 encoded text data. These are suitable for:
- Source code
- Configuration files
- Configuration files
- Log files
- JSON/XML data
- Plain text
@ -66,37 +72,39 @@ Servers expose a list of concrete resources via the `resources/list` endpoint. E
```typescript
{
uri: string; // Unique identifier for the resource
name: string; // Human-readable name
uri: string; // Unique identifier for the resource
name: string; // Human-readable name
description?: string; // Optional description
mimeType?: string; // Optional MIME type
mimeType?: string; // Optional MIME type
}
```
### Resource templates
### Resource templates
For dynamic resources, servers can expose URI templates that clients can use to construct valid resource URIs:
For dynamic resources, servers can expose [URI templates](https://datatracker.ietf.org/doc/html/rfc6570) that clients can use to construct valid resource URIs:
```typescript
{
uriTemplate: string; // URI template following RFC 6570
name: string; // Human-readable name for this type
description?: string; // Optional description
mimeType?: string; // Optional MIME type for all matching resources
mimeType?: string; // Optional MIME type for all matching resources
}
```
## Reading resources
To read a resource, clients make a `resources/read` request with the resource URI. The server responds with:
To read a resource, clients make a `resources/read` request with the resource URI.
The server responds with a list of resource contents:
```typescript
{
contents: [
{
uri: string; // The requested URI
uri: string; // The URI of the resource
mimeType?: string; // Optional MIME type
// One of:
text?: string; // For text resources
blob?: string; // For binary resources (base64 encoded)
@ -105,6 +113,10 @@ To read a resource, clients make a `resources/read` request with the resource UR
}
```
<Tip>
Servers may return multiple resources in response to one `resources/read` request. This could be used, for example, to return a list of files inside a directory when the directory is read.
</Tip>
## Resource updates
MCP supports real-time updates for resources through two mechanisms:
@ -113,7 +125,7 @@ MCP supports real-time updates for resources through two mechanisms:
Servers can notify clients when their list of available resources changes via the `notifications/resources/list_changed` notification.
### Content changes
### Content changes
Clients can subscribe to updates for specific resources:
@ -126,45 +138,58 @@ Clients can subscribe to updates for specific resources:
Here's a simple example of implementing resource support in an MCP server:
```typescript
const server = new Server({
name: "example-server",
version: "1.0.0"
});
// List available resources
server.setRequestHandler(ListResourcesRequestSchema, async () => {
return {
resources: [
{
uri: "file:///logs/app.log",
name: "Application Logs",
mimeType: "text/plain"
<Tabs>
<Tab title="TypeScript">
```typescript
const server = new Server({
name: "example-server",
version: "1.0.0"
}, {
capabilities: {
resources: {}
}
]
};
});
});
// Read resource contents
server.setRequestHandler(ReadResourceRequestSchema, async (request) => {
const uri = request.params.uri;
if (uri === "file:///logs/app.log") {
const logContents = await readLogFile();
return {
contents: [
{
uri,
mimeType: "text/plain",
text: logContents
}
]
};
}
throw new Error("Resource not found");
});
```
// List available resources
server.setRequestHandler(ListResourcesRequestSchema, async () => {
return {
resources: [
{
uri: "file:///logs/app.log",
name: "Application Logs",
mimeType: "text/plain"
}
]
};
});
// Read resource contents
server.setRequestHandler(ReadResourceRequestSchema, async (request) => {
const uri = request.params.uri;
if (uri === "file:///logs/app.log") {
const logContents = await readLogFile();
return {
contents: [
{
uri,
mimeType: "text/plain",
text: logContents
}
]
};
}
throw new Error("Resource not found");
});
```
</Tab>
<Tab title="Python">
```python
# Python implementation coming soon
```
</Tab>
</Tabs>
## Best practices
@ -194,4 +219,4 @@ When exposing resources:
- Encrypt sensitive data in transit
- Validate MIME types
- Implement timeouts for long-running reads
- Handle resource cleanup appropriately
- Handle resource cleanup appropriately

76
docs/concepts/sampling.mdx
View file

@ -5,6 +5,10 @@ description: "Let your servers request completions from LLMs"
Sampling is a powerful MCP feature that allows servers to request LLM completions through the client, enabling sophisticated agentic behaviors while maintaining security and privacy.
<Info>
This feature of MCP is not yet supported in the Claude Desktop client.
</Info>
## How sampling works
The sampling flow follows these steps:
@ -12,7 +16,7 @@ The sampling flow follows these steps:
1. Server sends a `sampling/createMessage` request to the client
2. Client reviews the request and can modify it
3. Client samples from an LLM
4. Client reviews the completion
4. Client reviews the completion
5. Client returns the result to the server
This human-in-the-loop design ensures users maintain control over what the LLM sees and generates.
@ -28,14 +32,24 @@ Sampling requests use a standardized message format:
role: "user" | "assistant",
content: {
type: "text" | "image",
// For text:
text?: string,
// For images:
data?: string, // base64 encoded
data?: string, // base64 encoded
mimeType?: string
}
}
],
modelPreferences?: {
hints?: [{
name?: string // Suggested model name/family
}],
costPriority?: number, // 0-1, importance of minimizing cost
speedPriority?: number, // 0-1, importance of low latency
intelligencePriority?: number // 0-1, importance of capabilities
},
systemPrompt?: string,
includeContext?: "none" | "thisServer" | "allServers",
temperature?: number,
@ -51,11 +65,27 @@ Sampling requests use a standardized message format:
The `messages` array contains the conversation history to send to the LLM. Each message has:
- `role`: Either "user" or "assistant"
- `role`: Either "user" or "assistant"
- `content`: The message content, which can be:
- Text content with a `text` field
- Image content with `data` (base64) and `mimeType` fields
### Model preferences
The `modelPreferences` object allows servers to specify their model selection preferences:
- `hints`: Array of model name suggestions that clients can use to select an appropriate model:
- `name`: String that can match full or partial model names (e.g. "claude-3", "sonnet")
- Clients may map hints to equivalent models from different providers
- Multiple hints are evaluated in preference order
- Priority values (0-1 normalized):
- `costPriority`: Importance of minimizing costs
- `speedPriority`: Importance of low latency response
- `intelligencePriority`: Importance of advanced model capabilities
Clients make the final model selection based on these preferences and their available models.
### System prompt
An optional `systemPrompt` field allows servers to request a specific system prompt. The client may modify or ignore this.
@ -85,8 +115,8 @@ The client returns a completion result:
```typescript
{
model: string, // Name of the model used
stopReason: "endTurn" | "stopSequence" | "maxTokens",
model: string, // Name of the model used
stopReason?: "endTurn" | "stopSequence" | "maxTokens" | string,
role: "user" | "assistant",
content: {
type: "text" | "image",
@ -97,33 +127,27 @@ The client returns a completion result:
}
```
## Example implementation
## Example request
Here's an example of requesting sampling from a client:
```typescript
// Server-side code
const response = await server.request({
method: "sampling/createMessage",
params: {
messages: [
```json
{
"method": "sampling/createMessage",
"params": {
"messages": [
{
role: "user",
content: {
type: "text",
text: "What files are in the current directory?"
"role": "user",
"content": {
"type": "text",
"text": "What files are in the current directory?"
}
}
],
systemPrompt: "You are a helpful file system assistant.",
includeContext: "thisServer",
maxTokens: 100,
temperature: 0.7
"systemPrompt": "You are a helpful file system assistant.",
"includeContext": "thisServer",
"maxTokens": 100
}
}, CreateMessageResultSchema);
// Handle the completion
console.log(response.content.text);
}
```
## Best practices
@ -218,4 +242,4 @@ Be aware of these limitations:
- Costs should be considered
- Model availability varies
- Response times vary
- Not all content types supported
- Not all content types supported

221
docs/concepts/tools.mdx
View file

@ -5,14 +5,19 @@ description: "Enable LLMs to perform actions through your server"
Tools are a powerful primitive in the Model Context Protocol (MCP) that enable servers to expose executable functionality to clients. Through tools, LLMs can interact with external systems, perform computations, and take actions in the real world.
## How tools work
<Note>
Tools are designed to be **model-controlled**, meaning that tools are exposed from servers to clients with the intention of the AI model being able to automatically invoke them (with a human in the loop to grant approval).
</Note>
Tools in MCP follow a request-response pattern where:
## Overview
1. Clients discover available tools through the `tools/list` endpoint
2. Clients can invoke tools using the `tools/call` endpoint
3. Servers execute the requested operation and return results
4. Progress updates can be sent during long-running operations
Tools in MCP allow servers to expose executable functions that can be invoked by clients and used by LLMs to perform actions. Key aspects of tools include:
- **Discovery**: Clients can list available tools through the `tools/list` endpoint
- **Invocation**: Tools are called using the `tools/call` endpoint, where servers perform the requested operation and return results
- **Flexibility**: Tools can range from simple calculations to complex API interactions
Like [resources](/docs/concepts/resources), tools are identified by unique names and can include descriptions to guide their usage. However, unlike resources, tools represent dynamic operations that can modify state or interact with external systems.
## Tool definition structure
@ -20,13 +25,11 @@ Each tool is defined with the following structure:
```typescript
{
name: string; // Unique identifier for the tool
name: string; // Unique identifier for the tool
description?: string; // Human-readable description
inputSchema: { // JSON Schema for the tool's parameters
inputSchema: { // JSON Schema for the tool's parameters
type: "object",
properties: {
// Tool-specific parameters
}
properties: { ... } // Tool-specific parameters
}
}
```
@ -35,73 +38,58 @@ Each tool is defined with the following structure:
Here's an example of implementing a basic tool in an MCP server:
```typescript
const server = new Server({
name: "example-server",
version: "1.0.0"
});
// Define available tools
server.setRequestHandler(ListToolsRequestSchema, async () => {
return {
tools: [{
name: "calculate_sum",
description: "Add two numbers together",
inputSchema: {
type: "object",
properties: {
a: { type: "number" },
b: { type: "number" }
},
required: ["a", "b"]
<Tabs>
<Tab title="TypeScript">
```typescript
const server = new Server({
name: "example-server",
version: "1.0.0"
}, {
capabilities: {
tools: {}
}
}]
};
});
});
// Handle tool execution
server.setRequestHandler(CallToolRequestSchema, async (request) => {
if (request.params.name === "calculate_sum") {
const { a, b } = request.params.arguments;
return {
toolResult: a + b
};
}
throw new Error("Tool not found");
});
```
// Define available tools
server.setRequestHandler(ListToolsRequestSchema, async () => {
return {
tools: [{
name: "calculate_sum",
description: "Add two numbers together",
inputSchema: {
type: "object",
properties: {
a: { type: "number" },
b: { type: "number" }
},
required: ["a", "b"]
}
}]
};
});
## Progress reporting
// Handle tool execution
server.setRequestHandler(CallToolRequestSchema, async (request) => {
if (request.params.name === "calculate_sum") {
const { a, b } = request.params.arguments;
return {
toolResult: a + b
};
}
throw new Error("Tool not found");
});
```
</Tab>
<Tab title="Python">
```python
# Python implementation coming soon
```
</Tab>
</Tabs>
For long-running operations, tools can report progress using the progress notification system:
## Example tool patterns
```typescript
server.setRequestHandler(CallToolRequestSchema, async (request) => {
if (request.params.name === "long_process") {
const progressToken = request.params._meta?.progressToken;
for (let i = 0; i < total; i++) {
await doWork();
// Send progress update
await server.notification({
method: "notifications/progress",
params: {
progress: i,
total: total,
progressToken
}
});
}
return { toolResult: "Complete" };
}
});
```
## Common tool patterns
Here are some common patterns for implementing tools:
Here are some examples of types of tools that a server could provide:
### System operations
@ -121,7 +109,7 @@ Tools that interact with the local system:
}
```
### API integrations
### API integrations
Tools that wrap external APIs:
@ -152,10 +140,10 @@ Tools that transform or analyze data:
type: "object",
properties: {
filepath: { type: "string" },
operations: {
type: "array",
items: {
enum: ["sum", "average", "count"]
operations: {
type: "array",
items: {
enum: ["sum", "average", "count"]
}
}
}
@ -169,7 +157,7 @@ When implementing tools:
1. Provide clear, descriptive names and descriptions
2. Use detailed JSON Schema definitions for parameters
3. Include examples in tool descriptions when helpful
3. Include examples in tool descriptions to demonstrate how the model should use them
4. Implement proper error handling and validation
5. Use progress reporting for long operations
6. Keep tool operations focused and atomic
@ -215,65 +203,48 @@ MCP supports dynamic tool discovery:
3. Tools can be added or removed during runtime
4. Tool definitions can be updated (though this should be done carefully)
## Common tool categories
Tools typically fall into these categories:
### System integration
- File operations
- Process management
- System information
- Network operations
### External services
- API calls
- Database operations
- Authentication services
- Cloud services
### Data processing
- Format conversion
- Data analysis
- Content generation
- Media processing
### Device control
- Hardware interaction
- Device configuration
- Sensor reading
- Output control
## Error handling
Tools should return clear error information:
Tool errors should be reported within the result object, not as MCP protocol-level errors. This allows the LLM to see and potentially handle the error. When a tool encounters an error:
1. Set `isError` to `true` in the result
2. Include error details in the `content` array
Here's an example of proper error handling for tools:
```typescript
try {
// Tool operation
const result = performOperation();
return {
content: [
{
type: "text",
text: `Operation successful: ${result}`
}
]
};
} catch (error) {
return {
toolResult: {
error: {
code: "OPERATION_FAILED",
message: "Failed to process request",
details: error.message
isError: true,
content: [
{
type: "text",
text: `Error: ${error.message}`
}
}
]
};
}
```
This approach allows the LLM to see that an error occurred and potentially take corrective action or request human intervention.
## Testing tools
Consider these testing approaches:
A comprehensive testing strategy for MCP tools should cover:
1. Unit test parameter validation
2. Mock external dependencies
3. Test progress reporting
4. Verify error handling
5. Test concurrent execution
6. Validate return values
7. Test rate limiting
8. Check resource cleanup
9. Test timeout handling
10. Verify security constraints
- **Functional testing**: Verify tools execute correctly with valid inputs and handle invalid inputs appropriately
- **Integration testing**: Test tool interaction with external systems using both real and mocked dependencies
- **Security testing**: Validate authentication, authorization, input sanitization, and rate limiting
- **Performance testing**: Check behavior under load, timeout handling, and resource cleanup
- **Error handling**: Ensure tools properly report errors through the MCP protocol and clean up resources

335
docs/concepts/transports.mdx
View file

@ -5,128 +5,13 @@ description: "Learn about MCP's communication mechanisms"
Transports in the Model Context Protocol (MCP) provide the foundation for communication between clients and servers. A transport handles the underlying mechanics of how messages are sent and received.
## How transports work
## Message Format
An MCP transport is built on a simple message-passing architecture where both clients and servers can send messages to each other. The transport layer is responsible for:
MCP uses [JSON-RPC](https://www.jsonrpc.org/) 2.0 as its wire format. The transport layer is responsible for converting MCP protocol messages into JSON-RPC format for transmission and converting received JSON-RPC messages back into MCP protocol messages.
- Establishing connections
- Sending messages
- Receiving messages
- Handling connection lifecycles
- Managing errors
- Ensuring reliable delivery
## Available transport types
MCP supports three main transport types:
### Standard input/output (stdio)
The stdio transport enables communication through standard input and output streams. This is particularly useful for local integrations and command-line tools.
```typescript
// Server implementation
const server = new Server({
name: "example-server",
version: "1.0.0"
});
const transport = new StdioServerTransport();
await server.connect(transport);
// Client implementation
const client = new Client({
name: "example-client",
version: "1.0.0"
});
const transport = new StdioClientTransport({
command: "./server",
args: ["--option", "value"]
});
await client.connect(transport);
```
### WebSocket
WebSocket transport provides real-time, bidirectional communication over a persistent connection. This is ideal for web-based applications that need low-latency messaging.
```typescript
// Server implementation
const server = new Server({
name: "example-server",
version: "1.0.0"
});
const transport = new WebSocketServerTransport();
await server.connect(transport);
// Client implementation
const client = new Client({
name: "example-client",
version: "1.0.0"
});
const transport = new WebSocketClientTransport(
new URL("ws://localhost:3000")
);
await client.connect(transport);
```
### Server-sent events
SSE transport enables server-to-client streaming with HTTP POST requests for client-to-server communication. This is useful when WebSocket support isn't available.
```typescript
// Server implementation
const server = new Server({
name: "example-server",
version: "1.0.0"
});
const transport = new SSEServerTransport("/message", response);
await server.connect(transport);
// Client implementation
const client = new Client({
name: "example-client",
version: "1.0.0"
});
const transport = new SSEClientTransport(
new URL("http://localhost:3000/sse")
);
await client.connect(transport);
```
## Transport interface
All transport implementations must conform to the Transport interface:
```typescript
interface Transport {
// Start processing messages
start(): Promise<void>;
// Send a JSON-RPC message
send(message: JSONRPCMessage): Promise<void>;
// Close the connection
close(): Promise<void>;
// Callbacks
onclose?: () => void;
onerror?: (error: Error) => void;
onmessage?: (message: JSONRPCMessage) => void;
}
```
## Message format
Messages sent over MCP transport follow the JSON-RPC 2.0 specification. There are three types of messages:
There are three types of JSON-RPC messages used:
### Requests
```typescript
{
jsonrpc: "2.0",
@ -137,7 +22,6 @@ Messages sent over MCP transport follow the JSON-RPC 2.0 specification. There ar
```
### Responses
```typescript
{
jsonrpc: "2.0",
@ -152,7 +36,6 @@ Messages sent over MCP transport follow the JSON-RPC 2.0 specification. There ar
```
### Notifications
```typescript
{
jsonrpc: "2.0",
@ -161,7 +44,130 @@ Messages sent over MCP transport follow the JSON-RPC 2.0 specification. There ar
}
```
## Error handling
## Built-in Transport Types
MCP includes two standard transport implementations:
### Standard Input/Output (stdio)
The stdio transport enables communication through standard input and output streams. This is particularly useful for local integrations and command-line tools.
Use stdio when:
- Building command-line tools
- Implementing local integrations
- Needing simple process communication
- Working with shell scripts
<Tabs>
<Tab title="TypeScript">
```typescript
// Server implementation
const server = new Server({
name: "example-server",
version: "1.0.0"
});
const transport = new StdioServerTransport();
await server.connect(transport);
// Client implementation
const client = new Client({
name: "example-client",
version: "1.0.0"
});
const transport = new StdioClientTransport({
command: "./server",
args: ["--option", "value"]
});
await client.connect(transport);
```
</Tab>
<Tab title="Python">
```python
# Python implementation coming soon
```
</Tab>
</Tabs>
### Server-Sent Events (SSE)
SSE transport enables server-to-client streaming with HTTP POST requests for client-to-server communication.
Use SSE when:
- Only server-to-client streaming is needed
- Working with restricted networks
- Implementing simple updates
<Tabs>
<Tab title="TypeScript">
```typescript
// Server implementation
const server = new Server({
name: "example-server",
version: "1.0.0"
});
const transport = new SSEServerTransport("/message", response);
await server.connect(transport);
// Client implementation
const client = new Client({
name: "example-client",
version: "1.0.0"
});
const transport = new SSEClientTransport(
new URL("http://localhost:3000/sse")
);
await client.connect(transport);
```
</Tab>
<Tab title="Python">
```python
# Python implementation coming soon
```
</Tab>
</Tabs>
## Custom Transports
MCP makes it easy to implement custom transports for specific needs. Any transport implementation just needs to conform to the Transport interface:
You can implement custom transports for:
- Custom network protocols
- Specialized communication channels
- Integration with existing systems
- Performance optimization
<Tabs>
<Tab title="TypeScript">
```typescript
interface Transport {
// Start processing messages
start(): Promise<void>;
// Send a JSON-RPC message
send(message: JSONRPCMessage): Promise<void>;
// Close the connection
close(): Promise<void>;
// Callbacks
onclose?: () => void;
onerror?: (error: Error) => void;
onmessage?: (message: JSONRPCMessage) => void;
}
```
</Tab>
<Tab title="Python">
```python
# Python implementation coming soon
```
</Tab>
</Tabs>
## Error Handling
Transport implementations should handle various error scenarios:
@ -173,29 +179,38 @@ Transport implementations should handle various error scenarios:
Example error handling:
```typescript
class ExampleTransport implements Transport {
async start() {
try {
// Connection logic
} catch (error) {
this.onerror?.(new Error(`Failed to connect: ${error}`));
throw error;
}
}
<Tabs>
<Tab title="TypeScript">
```typescript
class ExampleTransport implements Transport {
async start() {
try {
// Connection logic
} catch (error) {
this.onerror?.(new Error(`Failed to connect: ${error}`));
throw error;
}
}
async send(message: JSONRPCMessage) {
try {
// Sending logic
} catch (error) {
this.onerror?.(new Error(`Failed to send message: ${error}`));
throw error;
async send(message: JSONRPCMessage) {
try {
// Sending logic
} catch (error) {
this.onerror?.(new Error(`Failed to send message: ${error}`));
throw error;
}
}
}
}
}
```
```
</Tab>
<Tab title="Python">
```python
# Python implementation coming soon
```
</Tab>
</Tabs>
## Best practices
## Best Practices
When implementing or using MCP transport:
@ -210,34 +225,31 @@ When implementing or using MCP transport:
9. Monitor connection health
10. Implement proper security measures
## Security considerations
## Security Considerations
When implementing transport:
### Authentication and authorization
### Authentication and Authorization
- Implement proper authentication mechanisms
- Validate client credentials
- Use secure token handling
- Implement authorization checks
### Data security
### Data Security
- Use TLS for network transport
- Encrypt sensitive data
- Validate message integrity
- Implement message size limits
- Sanitize input data
### Network security
### Network Security
- Implement rate limiting
- Use appropriate timeouts
- Handle denial of service scenarios
- Monitor for unusual patterns
- Implement proper firewall rules
## Debugging transport
## Debugging Transport
Tips for debugging transport issues:
@ -251,28 +263,3 @@ Tips for debugging transport issues:
8. Monitor resource usage
9. Test edge cases
10. Use proper error tracking
## Transport selection guide
Choose the appropriate transport based on your needs:
### Use stdio when
- Building command-line tools
- Implementing local integrations
- Needing simple process communication
- Working with shell scripts
### Use WebSocket when
- Real-time updates are needed
- Low latency is required
- Bidirectional communication is necessary
- Working with web applications
### Use SSE when
- Only server-to-client streaming is needed
- WebSocket support is limited
- Working with restricted networks
- Implementing simple updates