How real projects brief Claude Code: the commands, conventions and warnings they put in CLAUDE.md.
nailuoGGCLAUDE.mdnailuoGG/anki-mcp-server/CLAUDE.md
when working with code in this repository. ## Repository Overview This is an MCP (Model Context Protocol) server that enables LLMs to interact with Anki flashcard software through AnkiConnect. The server
tinyhumansaiCLAUDE.mdtinyhumansai/opencompany/CLAUDE.md
feature `openhuman`) - `src/hive/`: hive desks — one `OpenHumanHive` per `[[group_chat]]`, speech over the `opencompany` MCP server, Jev routing, referral (`docs/spec/runtime/hive.md`) - `src/tiny/`: optional TinyAgents crate feature/status surface The command-line entrypoint
danielvm-gitCLAUDE.mddanielvm-git/bigpowers/CLAUDE.md
should prefer each one in the situations below. ### Preferred tools (MCP) The `sqz-mcp` server is registered in this project's MCP config. It exposes three read-only tools that
talkstreamCLAUDE.mdtalkstream/ru-text/.claude/CLAUDE.md
Notion AI Custom Skill template page in `notion/`, (2) MCP bridge (Claude Code + Notion MCP server). Notion AI Skills require Business/Enterprise plan. Template must be self-contained (no `${CLAUDE_PLUGIN
DMontgomery40CLAUDE.mdDMontgomery40/mcp-3D-printer-server/CLAUDE.md
testing. - `dist/` is gitignored in this repo (unlike bambu-printer-mcp). ## Architecture - Multi-printer MCP server: OctoPrint, Klipper, Duet, Repetier, Bambu, Prusa Connect, Creality. - `PRINTER_TYPE` env var selects
SentryCLAUDE.mdgetsentry/dotagents/skills/dotagents-qa/references/claude.md
requires one plugin skill, zero leaked client-extension agents, and both non-empty portable MCP servers. Claude Code 2.1.x rejects an `agents` field in plugin manifests. dotagents therefore omits
FarhanAliRazaCLAUDE.mdFarhanAliRaza/claude-context-local/CLAUDE.md
tests/run_tests.py --embeddings # Embedding tests only python tests/run_tests.py --search # Search tests only python tests/run_tests.py --mcp # MCP server tests only # Run tests with coverage python tests/run_tests.py --coverage # Run tests with verbose output
fazxesCLAUDE.mdfazxes/Claude-code/CLAUDE.md
resume via `--resume` (`PermissionRule.ts`, `coreSchemas.ts`, `hooks.ts`) - **`MCP_CONNECTION_NONBLOCKING=true` env var** — bounds MCP server connections at 5s instead of 30s (`client.ts`) - **Autocompact thrash loop** — actionable error message when circuit
MicrosoftCLAUDE.mdmicrosoft/Dataverse-skills/CLAUDE.md
files (`SKILL.md`) that teach the agent how to use the Python SDK, PAC CLI, MCP server, and Dataverse Web API. The plugin is loaded via: ```bash claude --plugin-dir .github/plugins/dataverse
ramakayCLAUDE.mdramakay/claude-self-reflect/CLAUDE.md
Single Rust binary (`csr-engine`). No Python, no Docker, no Qdrant. ``` csr-engine (44MB) ├── MCP server (rmcp, 15 tools) ├── Embeddings (FastEmbed, 384-dim, local) ├── Search (HNSW, <1ms p95) ├── Storage (SQLite
receptronCLAUDE.mdreceptron/mulmoterminal/CLAUDE.md
breaks the pattern.** It reads neither a flag nor a file in the directory: MCP servers are declared by an installed PLUGIN (`server/agents/muse-mcp.ts`), and `muse plugins install` records
JameZUKCLAUDE.mdJameZUK/Arkana/CLAUDE.md
Arkana Development Guide ## What is Arkana? Arkana is a Model Context Protocol (MCP) server exposing 308 binary analysis tools to AI clients. It supports PE, ELF, and Mach-O formats
timothywarner-orgCLAUDE.mdtimothywarner-org/claude-code/segment_2_context/CLAUDE.md
This `CLAUDE.md` itself, as a live teaching artifact ## What does NOT belong here - **No MCP-server-building topics.** MCP moved to `segment_4_hero
timothywarner-orgCLAUDE.mdtimothywarner-org/claude-code/segment_4_hero/CLAUDE.md
mcp.json` (repo root) is the canonical demo target. Do **not** write a new MCP server in this segment unless the learner specifically opts in to the optional homework. Point them
initializCLAUDE.mdinitializ/forge/CLAUDE.md
bearer (omitting it → 401 "missing org-id header"). Pattern lives in `admission_loader.go` / `remote_session_store.go` / `mcp/platform_token.go`. - **MCP auth types** (`auth.type`): `oauth`/`bearer`/`static` + managed `platform` (agent-principal token from `ForgeConfig.Platform.token_endpoint
netboxlabsCLAUDE.mdnetboxlabs/netbox-mcp-server/CLAUDE.md
NetBox MCP Server ## Core Concept A read-only [Model Context Protocol](https://modelcontextprotocol.io/) server that enables LLMs to interact with NetBox infrastructure data. Built with FastMCP and designed
KonghaYaoCLAUDE.mdKonghaYao/peri/peri-acp/CLAUDE.md
resume` / `fork`）里的 `mcpServers` acp 型声明在响应写入后由 `host/requests/acp_mcp.rs` 受理（`attach_session_servers`）；`mcp/connect` 只带 client 声明的 `serverId`，因此受理顺序不能提前到响应之前。会话级服务持有连接（每个会话一个 MCP 池），入站 `mcp/message` 按 `connectionId` 定位承载会话、未知连接返回 `-32001`，内层 MCP
KonghaYaoCLAUDE.mdKonghaYao/peri/peri-middlewares/CLAUDE.md
close_session`（幂等，`mcp/disconnect` 有上界）（ARC-MCP-ACP-001）。 - **System MCP 启动准入**：`system_mcp = true` 的 server 必须在首个 Reason 前完成 transport / initialize / 能力协商 / 真实 `tools/list`（空数组是成功结果，`Err` 不是发现证据），由启动闸门 hook
AppGramCLAUDE.mdAppGram/agentnotch/CLAUDE.md
watching │ ├── Settings/ │ │ └── AppSettings.swift # @AppStorage settings │ ├── Telemetry/ │ │ ├── TelemetryCoordinator.swift # OTLP telemetry handling │ │ └── OTLPHTTPServer.swift # HTTP server for telemetry │ └── Process/ │ └── MCPProcessManager.swift # MCP process management ├── Models/ │ ├── ClaudeCodeModels.swift # ClaudeSession, ClaudeToolExecution, etc. │ └── CodexModels.swift # CodexSession, CodexToolExecution, etc. ├── Views
ItamarZand88CLAUDE.mdItamarZand88/CLI-Anything-WEB/CLAUDE.md
generated CLIs 3. **FALLBACK**: `mcp__chrome-devtools__*` — only if playwright unavailable 4. **NEVER**: `mcp__claude-in-chrome__*` — cannot capture request bodies ## Generated CLI Structure Every generated CLI follows this exact
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.
