How real projects brief Codex, Cursor and every other agent that reads AGENTS.md.
nodetool-aiAGENTS.mdnodetool-ai/nodetool/packages/AGENTS.md
upload - WebSocket endpoint (`/ws`) for streaming workflow execution (MsgPack serialization) - OpenAI-compatible endpoint (`/v1/chat/completions`) - MCP server integration - Health check endpoint (`/health`) ```bash PORT=7777 HOST=127.0.0.1 node packages/websocket/dist/server.js ``` ### `@nodetool
8beeeaaatAGENTS.md8beeeaaat/touchdesigner-mcp/AGENTS.md
Repository Guidelines ## Project Structure & Module Organization `src/` contains the TypeScript MCP server: shared logic is in `core/`, MCP features in `features/`, transport code in `transport/`, and TouchDesigner communication in `tdClient
OpenAIAGENTS.mdopenai/role-specific-plugins/plugins/data-analytics/AGENTS.md
build` for all widget bundles, `npm run typecheck` for TypeScript, and `npm test` for MCP server and widget contract coverage. The plugin package is materialized directly from this directory. After
Sayhi-bzbAGENTS.mdSayhi-bzb/Agent-HTML/agent-html/AGENTS.md
Forbidden Imports And Actions - Do not call Codex app-server, the filesystem, shell commands, MCP servers, or privileged host APIs from artifact code. - Do not import `@/app/*`, `apps/agent-html-app`, `@/agent-html/runtime/ui`, `renderAgentHtml
iannuttallAGENTS.mdiannuttall/seo/AGENTS.md
source for `seo`, a local-first TypeScript SEO CLI, library, router skill, and stdio MCP server. The public repository is `iannuttall/seo`. The public npm package is the unscoped `seo` package
nirholasAGENTS.mdnirholas/XActions/AGENTS.md
whole session: **do you shell out to the CLI, or do you load the MCP server?** --- ## Two lanes into the same engine XActions ships both lanes from one install
ooplesAGENTS.mdooples/token-optimizer-mcp/integrations/AGENTS.md
Codex, OpenCode, and other AGENTS.md-aware agents) This project provides a `token-optimizer` MCP server. Use a named optimizer tool only when that exact schema is visible in the current
dbt-labsAGENTS.mddbt-labs/dbt-charts/AGENTS.md
Cloud. A PR that adds business logic to a CLI command file or MCP server module is rejected. ### Module boundaries tach enforces: `dbt_charts.cli` may depend only on `dbt_charts.agent
YoanWaiAGENTS.mdYoanWai/agent-manager/AGENTS.md
already does, with a buffer name that carries the pid so the manager, the MCP server, and the CLI never swap text between panes. ## Concurrent sessions Multiple agent sessions share
MicrosoftAGENTS.mdmicrosoft/DebugMCP/AGENTS.md
MCP/SSE → DebugMCPServer → DebuggingHandler → DebuggingExecutor → VS Code Debug API ``` ### Key Components | Component | Responsibility | Docs | |-----------|----------------|------| | `DebugMCPServer` | MCP server, tool/resource registration | [docs/architecture/debugMCPServer.md](docs/architecture/debugMCPServer.md) | | `DebuggingHandler` | Operation orchestration, state change detection | [docs/architecture/debuggingHandler.md](docs/architecture/debuggingHandler.md) | | `DebuggingExecutor
alisaittekeAGENTS.mdalisaitteke/photoshop-mcp/AGENTS.md
README.md](README.md) | Human docs, install, example prompts | | [server.json](server.json) | MCP Registry metadata | | [src/core/server.ts](src/core/server.ts) | MCP server entry | | [src/tools/](src/tools/) | Tool implementations | | [src/prompts/](src/prompts/) | MCP prompt templates | | [CONTRIBUTING.md](CONTRIBUTING.md
pulseaiclubAGENTS.mdpulseaiclub/phi/AGENTS.md
ExtensionPost.** Don't bypass the permission gate when changing the executor. Don't put MCP server tool schemas on the model — only `mcp_list` / `mcp_inspect` / `mcp_call`. - **Extensions
CranotAGENTS.mdCranot/roam-code/AGENTS.md
command names (280 canonical + 7 aliases). mcp_server.py # FastMCP server (17 tools in core preset; 246 in `full`) + `roam mcp` CLI command mcp_extras/ # MCP-native enhancements: sampling, watcher, session, progress
hhszzzzAGENTS.mdhhszzzz/taibu/AGENTS.md
pnpm install pnpm build # 构建所有 packages + Next.js pnpm lint pnpm test # 自动先构建 packages/core、mcp、mcp-server，再跑全量测试 ``` > **注意**：`pnpm test` 会先执行 `pnpm -C packages/core build && pnpm -C packages/mcp build && pnpm
HashiCorpAGENTS.mdhashicorp/design-system/AGENTS.md
packages/mcp` package is a Model Context Protocol server that exposes design system knowledge — component APIs, tokens, and usage guidance — to MCP-capable clients over a stdio transport. #### Key commands ```bash
Th0rgalAGENTS.mdTh0rgal/sandboxed.sh/agents.md
Runs **per workspace** using the Claude CLI. - Configuration is written to each workspace: - `.claude/settings.local.json` (MCP servers + permissions) - `.claude/skills/ /SKILL.md` (native skills with YAML frontmatter) - `CLAUDE.md` (general workspace context) - Auth: when
VikashLoombaAGENTS.mdVikashLoomba/copilot-mcp/AGENTS.md
README extraction in `src/McpAgent.ts` (`readmeExtractionRequest`) - final install URI open in `src/McpAgent.ts` (`openMcpInstallUri`) ### 3) Official MCP Registry search + install - UI: - `web/src/components/SearchRegistryServers.tsx` - `web/src/components/RegistryServerCard.tsx` - payload builders in `web/src/utils/registryInstall.ts` - RPC: - `registrySearchType` - `installFromConfigType` - `installClaudeFromConfigType` - `installCodexFromConfigType
agentregistry-devAGENTS.mdagentregistry-dev/agentregistry/AGENTS.md
centralized registry for securely curating, discovering, deploying, and managing agentic infrastructure including MCP (Model Context Protocol) servers, agents, and skills. **Tech Stack:** - **Backend/CLI:** Go 1.26+ - **Database:** PostgreSQL (accessed
PV-BhatAGENTS.mdPV-Bhat/vibe-check-mcp-server/AGENTS.md
Agent Quickstart Vibe Check MCP is a lightweight oversight layer for AI agents. It exposes two tools: - **vibe_check** – prompts you with clarifying questions to prevent tunnel vision. - **vibe_learn
PV-BhatAGENTS.mdPV-Bhat/vibe-check-mcp-server/docs/AGENTS.md
Agent Quickstart Vibe Check MCP is a lightweight oversight layer for AI agents. It exposes two tools: - **vibe_check** – prompts you with clarifying questions to prevent tunnel vision. - **vibe_learn
An open format for instructions to coding agents, read by Codex, Cursor and others. Think of it as a README written for agents.
At the repository root, with more specific files in subdirectories. Agents read the one closest to the file they're editing.
Setup and test commands, code style, and the rules a new contributor would need to know.
Claude Code reads CLAUDE.md. A one-line CLAUDE.md that points at AGENTS.md covers both.
