How real projects brief Claude Code: the commands, conventions and warnings they put in CLAUDE.md.
nikolausmCLAUDE.mdnikolausm/imap-mcp-server/CLAUDE.md
conventions — lives in **@AGENTS.md**. Read it first. ## Claude-specific notes - This is an IMAP/SMTP **MCP server** (TypeScript, ESM). It uses **`imapflow`** for IMAP and **`nodemailer`** for SMTP. (It does
vmihalisCLAUDE.mdvmihalis/hacker-bob/CLAUDE.md
installer copies the skills, update command shim, agents, rules, hooks, knowledge, bypass tables, MCP runtime, and generated settings. It merges `.mcp.json` and `.claude/settings.json` instead of overwriting unrelated config. 4. After
damusixCLAUDE.mddamusix/atomic-claude/CLAUDE.md
testing | atomic-writing | | `docs/guides/evaluations.md` | Docker eval environment, testing setup | atomic-writing | | `docs/guides/code-intel-mcp.md` | code-intel MCP server setup, .mcp.json, tools | atomic-writing | | `docs/reference/workflow.md` | plan, implement, diagnose, ship lifecycle | atomic-writing | | `docs/reference/commands.md
byaan-aiCLAUDE.mdbyaan-ai/byaan/CLAUDE.md
management │ │ ├── services/ # API client wrappers │ │ └── components/ # Reusable UI components │ └── src-tauri/ # Rust/Tauri desktop shell ├── server/ # FastAPI backend │ ├── routers/ # API endpoints │ ├── services/ # Business logic (unified_agent.py is 66KB - AI orchestration) │ ├── models/ # SQLAlchemy
conorluddyCLAUDE.mdconorluddy/xc-mcp/CLAUDE.md
CLAUDE.md XC-MCP is an MCP server exposing `xcodebuild`, `simctl` and `idb` as 77 tools, built to keep Xcode's enormous output inside a model's context. Published
hauptsacheNetCLAUDE.mdhauptsacheNet/typo3-mcp-server/CLAUDE.md
This project is about developing the mcp_server typo3 extension, that allows someone without technical knowledge to edit content using LLM's. - Run the tests using `composer test`. Test must
pzfreoCLAUDE.mdpzfreo/build123d-mcp/CLAUDE.md
existing failures — if tests are failing, fix them. ## Running the server ``` uv tool run --python 3.12 build123d-mcp ``` Communicates over stdio (MCP SDK v2 `MCPServer`). When configuring an MCP client
rcb0727CLAUDE.mdrcb0727/powerplatform-mcp-docs/CLAUDE.md
Power Platform MCP Server — Claude Code Instructions ## The 6-Phase Flow Workflow **ALWAYS follow this workflow when users want to create, build, or modify a Power Automate flow:** ### Phase
2389-researchCLAUDE.md2389-research/claude-plugins/CLAUDE.md
system-maintenance | [2389-research/remote-system-maintenance](https://github.com/2389-research/remote-system-maintenance) | Linux system diagnostics and maintenance via SSH/tmux | ### MCP Servers | Server | Repo | Description | |--------|------|-------------| | agent-drugs | [2389-research/agent-drugs](https://github.com/2389-research/agent-drugs) | Digital drugs that modify
8enSmithCLAUDE.md8enSmith/mcp-open-library/CLAUDE.md
full** test suite, so commits are slow but pre-verified. ## Architecture An MCP stdio server wrapping the Open Library HTTP API. Three layers: **`src/index.ts`** — the `OpenLibraryServer` class. Builds
FlorianBruniauxCLAUDE.mdFlorianBruniaux/ccboard/CLAUDE.md
Located in `ccboard-tui/src/`: - **13 tabs**: Dashboard, Sessions, Config, Hooks, Agents/Capabilities, Costs, History, MCP, Analytics, Activity, Search, Plugins, Brain - **Key bindings**: `Tab`/`Shift+Tab` (nav tabs), `j/k` (nav lists
HappenmassCLAUDE.mdHappenmass/omux/CLAUDE.md
else `en-US`. - `logger.ts` — async file logger. - `mcp-config.ts` — builds per-agent MCP server configuration (per-agent MCP scoping). - `ulid.ts` — Crockford-alphabet ULID generator. ### Doctor — `src/doctor/` - `run.ts` — `runDoctor()` executes checks. - `formatter.ts
optaveCLAUDE.mdoptave/ops-codegraph-tool/CLAUDE.md
Leiden, BFS, shortest path, centrality), classifiers (role, risk), builders (dependency, structure, temporal) | | **`mcp/`** | **MCP server** | | `mcp/` | MCP server exposing graph queries to AI agents; single-repo by default, `--multi-repo
gyorgyshCLAUDE.mdgyorgysh/keepresso/CLAUDE.md
notification first. - **Bundled tools** under `Contents/Helpers`: the `keepresso` CLI and the `keepresso-mcp` stdio MCP server (both drive automation leases). A ready-to-install agent skill ships under `Contents/Resources/AgentSkill/keep-awake`. - **Localization
reality-openedCLAUDE.mdreality-opened/openreality/CLAUDE.md
Open Reality stack (`reality-opened/openreality`), three components in one repo: - `mcp/`: the MCP server (npm `openreality-mcp`). Developed HERE directly; an all-plugin cordis app. Read `mcp/CLAUDE.md` before changing
reality-openedCLAUDE.mdreality-opened/openreality/mcp/CLAUDE.md
public `reality-opened/openreality` monorepo (npm `openreality-mcp`): Open Reality OS as MCP tools. An **all-plugin [cordis](https://github.com/cordiverse/cordis) app**. Read the Architecture section of [README.md](README.md) before
clawdotnetCLAUDE.mdclawdotnet/SharpClawCode/CLAUDE.md
named scenarios | | `SharpClaw.Code.ParityHarness` | End-to-end scenarios via `ParityTestHost` with mock LLM | | `SharpClaw.Code.Mcp.FixtureServer` | MCP fixture server for testing | Mock provider scenarios: `streaming_text`, `stream_failure`, `stream_slow`, `read_file_roundtrip
hyperfx-aiCLAUDE.mdhyperfx-ai/marketing-skills/CLAUDE.md
against the live MCP catalog (the ` .json` schema files exposed by the connected MCP server). The MCP descriptor wins over codebase function names, prior skill content, or memory. - **Allowed Hyper
JeiKeiLimCLAUDE.mdJeiKeiLim/tenet/CLAUDE.md
CLIs (Claude Code, OpenCode, Codex) using a persistent SQLite state store, exposed via an MCP server and CLI. ## Build & Development Commands A `Makefile` wraps the common `pnpm` scripts — prefer
vouchdevCLAUDE.mdvouchdev/vouch/CLAUDE.md
manifest are silently ignored by current loaders — the sync test rejects them. The kb.* MCP server is deployment config (`openclaw mcp add vouch -- vouch serve`); the trust boundary lives
A file Claude Code reads at the start of every session. It holds the commands, conventions and warnings the agent needs for this project.
At the repository root. Claude Code also reads CLAUDE.md files in subdirectories when it works there.
Build and test commands, the project's layout, conventions that aren't obvious from the code, and mistakes to avoid. Short files tend to work better than long ones.
Claude Code reads CLAUDE.md; most other agents read AGENTS.md. Many projects keep one and point the other at it.
