Skip to main content

modelcontextprotocol

MCP specification and SDK reference links. Use when you need to look up MCP docs, spec sections, GitHub repos, or SDK links — NOT as general context for MCP development tasks.

Source facts

Repository
cyanheads/claude-config
Last source activity
October 10, 2026 at 21:10
Detected SKILL.md language
English
Stars
1
Forks
0

Install options

The review-first prompt is selected by default. You can switch to a direct command or download a local copy.

Review the source files

Read SKILL.md and any companion files shown by SkillsMP before deciding whether to install.

Showing SKILL.md

SKILL.md
Source instructions · Read-only preview
name
modelcontextprotocol
description
MCP specification and SDK reference links. Use when you need to look up MCP docs, spec sections, GitHub repos, or SDK links — NOT as general context for MCP development tasks.
user-invokable
true
# Model Context Protocol (MCP) Open standard for connecting AI applications to external data sources, tools, and workflows. JSON-RPC 2.0 based, with defined primitives and transport layers. Verified against modelcontextprotocol.io and the `modelcontextprotocol` GitHub org on 2026-10-10. ## Architecture ``` Host (AI app) └─ Client (1:1 with server) └─ Server (exposes primitives) ``` - **Host**: The AI application (Claude Desktop, IDE, agent framework) - **Client**: Connects a host to one server and supplies protocol version, identity, and capabilities on every request. - **Server**: Exposes tools, resources, and prompts over a transport. Servers can be **local** (stdio, launched as subprocess) or **remote** (Streamable HTTP). Authorization is optional in the spec; when used over HTTP it is OAuth 2.1-based. ## Primitives | Primitive | Offered by | Purpose | | --------------- | ---------- | ------------------------------------------------------------ | | **Tools** | Server | Functions the model can call (side effects, computation) | | **Resources** | Server | Data the client can attach as context | | **Prompts** | Server | Templated messages/workflows for specific tasks | | **Elicitation** | Client | Server asks the user for input via the client (form or URL mode) | | **Sampling** | Client | Server asks the client for an LLM completion. Deprecated in 2026-07-28 (SEP-2577). | | **Roots** | Client | Client tells the server which filesystem roots it can access. Deprecated in 2026-07-28 (SEP-2577). | In 2026-07-28, client features are reached through Multi Round-Trip Requests: the server returns an `input_required` result carrying `inputRequests`, and the client retries the original request with `inputResponses`. Earlier revisions used server-initiated requests (`elicitation/create`, `sampling/createMessage`, `roots/list`). ## Transports | Transport | Use Case | Notes | | ------------------- | -------------- | ---------------------------------------------------------- | | **stdio** | Local servers | Launched as subprocess, pipes for communication | | **Streamable HTTP** | Remote servers | Single endpoint; POST per request, response as JSON or a request-scoped SSE stream. 2026-07-28 removed `Mcp-Session-Id` sessions, the GET stream, and SSE resumability (`Last-Event-ID`); requires `Mcp-Method`/`Mcp-Name` headers. | The legacy HTTP+SSE transport (separate `/sse` + `/messages` endpoints) is Deprecated under the feature lifecycle policy; use Streamable HTTP. ## Request Model (2026-07-28) 1. **Per-request metadata**: Every request carries `io.modelcontextprotocol/protocolVersion` and `io.modelcontextprotocol/clientCapabilities` in `_meta`, and should carry `io.modelcontextprotocol/clientInfo`. Results should carry `io.modelcontextprotocol/serverInfo`. 2. **Optional discovery**: Servers must implement `server/discover` (supported versions, capabilities, identity); clients may call it first or just send requests and handle `UnsupportedProtocolVersionError`. 3. **Stateless operation**: No `initialize` handshake, sessions, or `ping`. Cross-call state uses explicit server-minted handles passed as tool arguments. Every result has `resultType` (`"complete"` or `"input_required"`). 4. **Change notifications**: Clients opt in through `subscriptions/listen`, a long-lived POST response stream (replaces `resources/subscribe` and the HTTP GET stream). Progress and log notifications stay on the originating request's stream; log level is set per request via `io.modelcontextprotocol/logLevel`. Revisions through `2025-11-25` use the `initialize`/`initialized` handshake. The 2026-07-28 versioning page covers backward compatibility with them. ## Capability Negotiation Clients declare capabilities on every request. Servers return theirs from `server/discover`. Only use features the other side declared. **Server capabilities**: `tools`, `resources`, `prompts`, `completions`, `logging` (deprecated in 2026-07-28, SEP-2577) **Client capabilities**: `elicitation` (`form`, `url`); `sampling` and `roots` are deprecated in 2026-07-28 (SEP-2577), and new implementations should not add them. Optional extensions are negotiated through the `extensions` map in both capability objects, keyed by identifier (`{vendor-prefix}/{name}`; official ones use `io.modelcontextprotocol/`). ## Specification Current revision: **2026-07-28** (`/specification/latest` redirects here). A draft at `/specification/draft` accumulates changes for the next revision; its changelog was empty as of 2026-10-10. Previous revisions: `2025-11-25`, `2025-06-18`, `2025-03-26`, `2024-11-05`. **2026-07-28 key changes**: stateless protocol (no handshake, sessions, or `ping`; SEP-2575, SEP-2567), mandatory `server/discover`, Multi Round-Trip Requests replacing server-initiated requests (SEP-2322), `subscriptions/listen`, required `ttlMs`/`cacheScope` on list and read results (SEP-2549), tasks moved to an extension (SEP-2663), any JSON Schema 2020-12 keyword allowed in `inputSchema`/`outputSchema` (SEP-2106), resource-not-found error code changed from `-32002` to `-32602`, and spec error codes moved into `-32020`..`-32099`. Deprecated: Roots, Sampling, Logging, Dynamic Client Registration (replaced by Client ID Metadata Documents), and HTTP+SSE. The full list is on the Changelog page. | Section | URL | | ------------------- | ----------------------------------------------------------------------------------------------------- | | Overview | https://modelcontextprotocol.io/specification/2026-07-28 | | Architecture | https://modelcontextprotocol.io/specification/2026-07-28/architecture | | Base Protocol | https://modelcontextprotocol.io/specification/2026-07-28/basic | | Versioning | https://modelcontextprotocol.io/specification/2026-07-28/basic/versioning | | Message Patterns | https://modelcontextprotocol.io/specification/2026-07-28/basic/patterns | | Multi Round-Trip | https://modelcontextprotocol.io/specification/2026-07-28/basic/patterns/mrtr | | Subscriptions | https://modelcontextprotocol.io/specification/2026-07-28/basic/patterns/subscriptions | | Cancellation | https://modelcontextprotocol.io/specification/2026-07-28/basic/patterns/cancellation | | Progress | https://modelcontextprotocol.io/specification/2026-07-28/basic/patterns/progress | | Transports | https://modelcontextprotocol.io/specification/2026-07-28/basic/transports | | stdio | https://modelcontextprotocol.io/specification/2026-07-28/basic/transports/stdio | | Streamable HTTP | https://modelcontextprotocol.io/specification/2026-07-28/basic/transports/streamable-http | | Authorization | https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization | | Client Registration | https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization/client-registration | | AS Discovery | https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization/authorization-server-discovery | | Auth Security | https://modelcontextprotocol.io/specification/2026-07-28/basic/authorization/security-considerations | | Server Features | https://modelcontextprotocol.io/specification/2026-07-28/server | | Discovery | https://modelcontextprotocol.io/specification/2026-07-28/server/discover | | Tools | https://modelcontextprotocol.io/specification/2026-07-28/server/tools | | Resources | https://modelcontextprotocol.io/specification/2026-07-28/server/resources | | Prompts | https://modelcontextprotocol.io/specification/2026-07-28/server/prompts | | Caching | https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/caching | | Completion | https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/completion | | Pagination | https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/pagination | | Logging | https://modelcontextprotocol.io/specification/2026-07-28/server/utilities/logging | | Elicitation | https://modelcontextprotocol.io/specification/2026-07-28/client/elicitation | | Sampling | https://modelcontextprotocol.io/specification/2026-07-28/client/sampling | | Roots | https://modelcontextprotocol.io/specification/2026-07-28/client/roots | | Deprecated Features | https://modelcontextprotocol.io/specification/2026-07-28/deprecated | | Changelog | https://modelcontextprotocol.io/specification/2026-07-28/changelog | | Schema Reference | https://modelcontextprotocol.io/specification/2026-07-28/schema | | Draft | https://modelcontextprotocol.io/specification/draft | The normative schema source is `schema/<revision>/schema.ts` in the `modelcontextprotocol/modelcontextprotocol` repo. ## Documentation Docs are versioned per revision (`/docs/<revision>/...`); the unversioned paths below redirect to the current revision. | Page | URL | | ----------------------- | ---------------------------------------------------------------------------------- | | Introduction | https://modelcontextprotocol.io/docs/getting-started/intro | | Architecture | https://modelcontextprotocol.io/docs/learn/architecture | | Server Concepts | https://modelcontextprotocol.io/docs/learn/server-concepts | | Client Concepts | https://modelcontextprotocol.io/docs/learn/client-concepts | | Versioning | https://modelcontextprotocol.io/docs/learn/versioning | | Build a Server | https://modelcontextprotocol.io/docs/develop/build-server | | Build a Client | https://modelcontextprotocol.io/docs/develop/build-client | | Client Best Practices | https://modelcontextprotocol.io/docs/develop/clients/client-best-practices | | Connect Local Servers | https://modelcontextprotocol.io/docs/develop/connect-local-servers | | Connect Remote Servers | https://modelcontextprotocol.io/docs/develop/connect-remote-servers | | Authorization Tutorial | https://modelcontextprotocol.io/docs/tutorials/security/authorization | | Security Best Practices | https://modelcontextprotocol.io/docs/tutorials/security/security_best_practices | | Local Server Security | https://modelcontextprotocol.io/docs/tutorials/security/local-server-security | | Inspector | https://modelcontextprotocol.io/docs/tools/inspector | | Debugging | https://modelcontextprotocol.io/docs/tools/debugging | | SDKs | https://modelcontextprotocol.io/docs/sdk | | Examples | https://modelcontextprotocol.io/examples | | LLM-friendly index | https://modelcontextprotocol.io/llms.txt (append `.md` to any page URL for raw Markdown) | ## Extensions, Registry, Governance | Page | URL | | ------------------------ | ------------------------------------------------------ | | Extensions Overview | https://modelcontextprotocol.io/extensions/overview | | Extension Support Matrix | https://modelcontextprotocol.io/extensions/client-matrix | | MCP Apps | https://modelcontextprotocol.io/extensions/apps/overview | | Tasks | https://modelcontextprotocol.io/extensions/tasks/overview | | Skills over MCP | https://modelcontextprotocol.io/extensions/skills/overview | | Registry (docs) | https://modelcontextprotocol.io/registry/about | | Registry publish quickstart | https://modelcontextprotocol.io/registry/quickstart | | Registry (live, preview) | https://registry.modelcontextprotocol.io/ | | SEP Index | https://modelcontextprotocol.io/seps | | SDK Tiers | https://modelcontextprotocol.io/community/sdk-tiers | | Feature Lifecycle | https://modelcontextprotocol.io/community/feature-lifecycle | | Roadmap | https://modelcontextprotocol.io/development/roadmap |
View on GitHub
This SKILL.md is very large, so SkillsMP previews the first section here. View on GitHub