MCP
Named by 45 kept items in this workspace.
Search every workspace for MCP. It matches the name as written, so it can find other things that share it.
The items that name it 45
Protocol through which kanzo exposes its board to coding agents.
Protocol that authenticates the channel, but the paper says the sender remains the party in doubt.
Tool protocol; a rogue MCP server was added to Codex's config in the reported attack.
Tool interface through which jshook exposes its tools to AI agents.
Model Context Protocol; farol reads MCP tools, resources and prompts in Python code as entry points.
Optional adapter translating MCP into PRAXO-CORE contracts.
Model Context Protocol; GenOffice exposes 29 MCP tools for agents.
Protocol for exposing tools to agents through servers. A commenter notes 207 tool schemas in context add a cost to every call.
Abstracts away implementation so different AI products can use the tools needed to do productive work.
Tool interface through which agents access the backend in each task.
Used device-side to control speakers, LEDs, servos and GPIO, and cloud-side to extend assistants to smart-home control, desktop actions, knowledge search and email.
First-class Alter integration; a guide shows how to build an MCP server that serves Alter-managed credentials as tools.
Protocol Salesforce is said to expose, with APIs, so customers plug its data into their own agents.
Protocol for giving the agent tools and context; used selectively
Protocol for connecting tool loops to tool servers; the post says building an MCP server was already easy, but integrating your own tool loop with an external MCP server had to be figured out by the developer
G-Krusch exposes 12 MCP tools over SSE and Stdio, plus a REST bridge on port 5446, for Drive, Calendar and Gmail.
Supported integration for Tracekit; Tracekit also provides an MCP gateway that runs calls with keys only it holds
Protocol over which Graphicle serves code-graph and memory tools to clients; graphicle mcp serves them
Model Context Protocol servers, one of the resource types discoverable via ARD
Model Context Protocol servers, one of the resource types ARD makes discoverable
Used to connect the user's AI subscription to PDF TO MP3 for document processing, via a Streamable HTTP server address
Model and MCP lists were moved to a separate catalog fetched at runtime so new models and plugins ship without a release.
Connects an AI agent to tools and data; a server exposes tools and resources while the agent application acts as the MCP host, keeping one client per server. Messages are transferred in JSON-RPC 2.0 format over stdio for local servers and Streamable HTTP for remote ones.
available inside the phone app's Alpine Linux/proot sandbox alongside a shell and a browser, for tool use
Protocol used to expose Codex's Computer Use (cua_repl) as a server so other agents like Claude Code can register and call it
Protocol for tool sessions; each task defines the available MCP tools and every attempt gets an isolated, freshly initialized MCP session
protocol by which agents call tools and capture traces of tool outputs and source IDs, which ProvenanceGuard reads to verify answers
Used e.g. as the artifact MCP so multiple Claudes can access shared artifact data, and for accessing shared or personal credentials/tools like Google Docs within Claude Tag.
Tool-use integration point for the project: initialize/tools/list stay free for discovery, while tools/call, resources/read and prompts/get are billed via the L402-EL2 payment flow
Protocol used to connect agents like Claude Code, Codex or Cursor to external tools/servers such as Google Flow, Adobe Premiere Pro, or After Effects for video work.
The gateway fronts MCP servers behind client keys and can run their tools inside chat requests for clients that do not support MCP. A merged /mcp endpoint is provided.
工具/服务器协议，笔记提到其认证是少数带真正复杂性的领域之一
Protocol over which every supported client starts the same bundled dist/server.js process locally via stdio, used to expose the jev_review tool to coding agents.
integration layer consolidating as the standard for tool/agent delegation; described as better than CLI for most custom-harness integrations
Protocol/tool interface through which EvoOntology exposes its versioned Ontology Layer as a universal agent plugin, letting agents call browse_semantics and resolve_semantics on demand.
Protocol that exposes every gr.Workflow output node as a tool an AI assistant (e.g. Claude Code, Cursor) can call, enabled via mcp_server=True.
mcp/server.py is a stdio JSON-RPC server exposing the tools; tools/list derived from the contract, core 12 listed by default, all 42 with FFMPEG_SKILL_MCP_FULL=1; also advertises five prompt recipes (reel, podcast, multicam, delivery_check, hdr)
Used in api-routes.json's question-to-API router, tried first before keyed and keyless APIs for live data retrieval.
Tool-use protocol covered within the agents section (05) alongside tool design and guardrails
Covered in the course as a way to give agents tool use, taught in comparison to using a plain API ("MCP vs API")
Protocol serena uses to force symbol-level retrieval of code instead of raw file reads/edits.
Protocol used to register CodeBurn as a local stdio server so an agent can query usage/savings; also the subject of CodeBurn's detection of unused/underused MCP servers whose tool schemas ride along in every prompt.
Protocol used to connect agents to real internal tools reliably and in a governable way.
Rikugan can integrate MCP servers and reuse those from Claude Code and Codex. It does not need an MCP server to reach the host database.
One of the integration categories OpenSRE connects to across the modern cloud stack, alongside LLMs, observability, infrastructure, databases, and incident management tools.
All technologies in this workspace.