diff --git a/.env.example b/.env.example
new file mode 100644
index 0000000..9829512
--- /dev/null
+++ b/.env.example
@@ -0,0 +1,37 @@
+# spark-mcp-go environment configuration
+# Copy to .env and edit the *REQUIRED* values.
+
+# HTTP listen address (default: :8080)
+LISTEN_ADDR=:8080
+
+# Root directory for SQLite, uploads, and logs (default: ./data)
+DATA_DIR=./data
+
+# *REQUIRED* Comma-separated admin bearer tokens for /admin/* endpoints
+ADMIN_TOKENS=change-me-admin-token-1,change-me-admin-token-2
+
+# *REQUIRED* Agent bearer token for /mcp endpoints
+AGENT_TOKEN=change-me-agent-token
+
+# Timeout for fetch_url and RM tool HTTP calls (default: 30s)
+HTTP_CLIENT_TIMEOUT=30s
+
+# Maximum HTTP response bytes before truncation (default: 1048576)
+MAX_RESPONSE_BYTES=1048576
+
+# Timeout for spark-submit process execution (default: 60s)
+SPARK_SUBMIT_TIMEOUT=60s
+
+# Log directory (default: ./data/logs)
+LOG_DIR=./data/logs
+
+# Log level: debug, info, warn, error (default: info)
+LOG_LEVEL=info
+
+# Log format: text (terminal) or json (log file; default: text)
+LOG_FORMAT=text
+
+# Analyzer heuristic thresholds (defaults shown)
+ANALYZER_DATA_SKEW_RATIO=3.0
+ANALYZER_GC_PRESSURE_RATIO=0.1
+ANALYZER_BOTTLENECK_SHUFFLE_GB=50
diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md
new file mode 100644
index 0000000..9eeb61d
--- /dev/null
+++ b/ARCHITECTURE.md
@@ -0,0 +1,63 @@
+# spark-mcp-go 架构说明
+
+## 包依赖图
+
+```
+main
+  ├─ config
+  ├─ logging
+  ├─ storage (SQLite)
+  ├─ admin
+  │    └─ storage / cluster / audit
+  ├─ mcp
+  │    ├─ tools
+  │    │    ├─ rm
+  │    │    ├─ httpclient
+  │    │    ├─ analyzer
+  │    │    └─ storage / cluster
+  │    └─ server (mark3labs/mcp-go)
+  └─ middleware
+```
+
+- `admin` 只依赖 `storage`、`cluster`、`audit`、`middleware`。
+- `tools` 依赖 `rm`（ResourceManager 客户端）、`httpclient`（通用 HTTP 原语）、`analyzer`（日志分析）。
+- `mcp` 聚合所有 Tools，注册到 `mark3labs/mcp-go` 的 Streamable HTTP server。
+- `main` 负责加载配置、打开数据库、挂载路由、启动 HTTP 服务。
+
+## 数据流
+
+1. 管理员调用 `/admin/clusters` POST，写入 cluster JSON。
+2. `storage` 把 cluster 持久化到 SQLite（后续会对敏感字段加密）。
+3. LLM Agent 调用 `/mcp` Tool（如 `list_clusters`）。
+4. Tool 从 storage 读取 cluster，携带 auth 信息，通过 `httpclient` 或本地 `spark-submit` 调用 YARN/Spark。
+5. 结果返回给 LLM；每次 Tool 调用同时写入独立审计日志。
+
+## 11 Tools 分类
+
+| 分类 | Tools |
+|---|---|
+| discovery | `list_clusters` |
+| exec | `spark_submit` |
+| RM | `list_applications`, `get_application_status`, `get_application_logs`, `kill_application`, `fetch_cluster_env` |
+| SHS | `fetch_spark_metrics` |
+| analyzer | `analyze_spark_log` |
+| 原语 | `fetch_url`, `upload_file` |
+
+## 关键设计决策
+
+- **slice-form `spark_submit`**：参数以字符串数组传入，避免 shell 拼接和注入。
+- **cluster `url_allowlist`**：所有外发 HTTP（`fetch_url`、RM、SHS）必须匹配 allowlist，防止 SSRF。
+- **日志降级链**：`get_application_logs` 依次尝试 `amContainerLogs`、aggregated-logs、通用 logs 端点。
+- **多 token admin**：`ADMIN_TOKENS` 支持逗号分隔多管理员 token，便于轮换。
+- **per-tool 独立审计日志**：每个 Tool 每次调用写入 `${LOG_DIR}/tools/<tool>.log`，文件权限 `0600`。
+- **`auth_password` 不可通过 JSON 创建**：字段带 `json:"-"` tag，避免接口泄露密码；生产通过专用端点或数据库初始化写入。
+
+## 安全模型（三道防线）
+
+1. **Token 鉴权**：`ADMIN_TOKENS` 保护管理面，`AGENT_TOKEN` 保护 MCP 面。
+2. **SSRF 防护**：`fetch_url` 与所有集群请求都走 cluster allowlist，且内置 DNS rebinding 检查。
+3. **路径/参数白名单**：`spark_submit` 仅允许 jar、class、--conf 等安全参数；禁止 shell 元字符与重定向。
+
+## 运行与部署
+
+详见 [docs/runbook.md](docs/runbook.md)。
diff --git a/README.md b/README.md
new file mode 100644
index 0000000..15ac266
--- /dev/null
+++ b/README.md
@@ -0,0 +1,139 @@
+# spark-mcp-go
+
+MCP (Model Context Protocol) server for Apache Spark on YARN. Lets an LLM agent
+discover Spark/YARN endpoints, submit jobs, fetch logs, and analyze them via
+11 typed Tools. Streamable HTTP transport, SQLite-backed configuration, and
+`log/slog` structured logging with per-Tool call files.
+
+## What it does
+
+- Exposes 11 MCP Tools over Streamable HTTP at `/mcp`
+- Admin API at `/admin/*` for cluster and audit configuration
+- HTTP Basic and YARN SimpleAuth for Spark/YARN endpoints
+- SSRF protection with cluster URL allowlist plus DNS rebinding guard
+- Per-Tool independent audit log (file mode `0600`)
+- `spark-submit` via local `exec.Command` (no shell, no injection)
+
+## Quick start (3 steps)
+
+1. Copy and edit the environment file:
+   ```bash
+   cp .env.example .env
+   # Edit ADMIN_TOKENS and AGENT_TOKEN
+   ```
+
+2. Build and start the server:
+   ```bash
+   go build ./...
+   ./spark-mcp-go
+   # Or use the helper:
+   # ./scripts/dev.sh
+   ```
+
+3. Initialize an MCP session, then call `list_clusters` to discover endpoints.
+
+## 11 Tools
+
+| Tool | Type | Purpose |
+|---|---|---|
+| `list_clusters` | discovery | Return all active clusters (LLM entry point) |
+| `spark_submit` | exec | Local `spark-submit` process, extracts `app_id` |
+| `list_applications` | RM | `GET /ws/v1/cluster/apps[?state=&user=]` |
+| `get_application_status` | RM | `GET /ws/v1/cluster/apps/{id}` |
+| `get_application_logs` | RM | Fallback chain: `amContainerLogs` -> aggregated-logs -> logs |
+| `kill_application` | RM | `PUT` state `KILLED` |
+| `fetch_spark_metrics` | SHS | Executor metrics + summary mode triggers analyzer |
+| `fetch_cluster_env` | RM | Aggregate `/cluster/info` and `/cluster/metrics` |
+| `analyze_spark_log` | analyzer | RM logs + heuristic rules + LLM-ready prompt |
+| `fetch_url` | HTTP | Generic primitive, uses cluster allowlist and auth |
+| `upload_file` | FS | Write to `./data/uploads/`, reference from `spark_submit` |
+
+## Environment variables
+
+| Variable | Default | Required | Description |
+|---|---|---|---|
+| `LISTEN_ADDR` | `:8080` | no | HTTP listen address |
+| `DATA_DIR` | `./data` | no | SQLite, uploads, and log root |
+| `ADMIN_TOKENS` | — | **yes** | Comma-separated tokens for `/admin/*` |
+| `AGENT_TOKEN` | — | **yes** | Token for `/mcp` requests |
+| `HTTP_CLIENT_TIMEOUT` | `30s` | no | Timeout for `fetch_url`/RM Tool HTTP calls |
+| `MAX_RESPONSE_BYTES` | `1048576` (1 MiB) | no | HTTP response truncation limit |
+| `SPARK_SUBMIT_TIMEOUT` | `60s` | no | `spark-submit` process timeout |
+| `LOG_DIR` | `./data/logs` | no | Log root directory |
+| `LOG_LEVEL` | `info` | no | `debug`, `info`, `warn`, or `error` |
+| `LOG_FORMAT` | `text` | no | `text` for terminal, `json` for log files |
+| `ANALYZER_DATA_SKEW_RATIO` | `3.0` | no | Data-skew rule threshold (max/min ratio) |
+| `ANALYZER_GC_PRESSURE_RATIO` | `0.1` | no | GC-pressure rule threshold (GC/CPU ratio) |
+| `ANALYZER_BOTTLENECK_SHUFFLE_GB` | `50` | no | Bottleneck rule threshold (shuffle GB) |
+
+## Admin API
+
+All endpoints require `Authorization: Bearer <admin_token>`.
+
+| Method | Path | Description |
+|---|---|---|
+| `GET` | `/admin/clusters` | List all clusters |
+| `POST` | `/admin/clusters` | Create a cluster |
+| `GET` | `/admin/clusters/:id` | Get one cluster |
+| `PUT` | `/admin/clusters/:id` | Update a cluster |
+| `DELETE` | `/admin/clusters/:id` | Delete a cluster |
+| `GET` | `/admin/audit?limit=100` | Query audit log |
+
+Create a cluster:
+
+```bash
+curl -sS -X POST \\
+  -H "Authorization: Bearer ${ADMIN_TOKEN}" \\
+  -H "Content-Type: application/json" \\
+  -d '{
+    "id": "prod",
+    "name": "prod",
+    "rm_url": "http://rm.example.com:8088",
+    "shs_url": "http://shs.example.com:18080",
+    "spark_submit_execute_bin": "/opt/spark/bin/spark-submit",
+    "is_active": true,
+    "auth_type": "simple",
+    "auth_username": "yarn",
+    "rate_limit_per_min": 10,
+    "url_allowlist": ["rm.example.com:8088", "shs.example.com:18080"]
+  }' \\
+  "http://127.0.0.1:${LISTEN_ADDR:-:8080}/admin/clusters"
+```
+
+Note: `auth_password` is not accepted via JSON (`json:"-"`). Set it through a
+dedicated password endpoint or seed the database directly.
+
+## Security model
+
+Three lines of defense:
+
+1. **Token auth** - `ADMIN_TOKENS` protects `/admin/*`; `AGENT_TOKEN` protects `/mcp`.
+2. **SSRF guard** - Every outbound URL must match the cluster's `url_allowlist`,
+   with private-IP CIDR blacklist (incl. `169.254.0.0/16` AWS/GCP metadata).
+3. **Path allowlist** - `upload_file` writes only to `./data/uploads/`, rejects
+   absolute paths, `..`, and non-`[a-zA-Z0-9._-]` filenames.
+
+Additional guarantees:
+
+- `spark_submit` runs the cluster's binary via `exec.Command(name, args...)` —
+  slice form, never `sh -c`. No shell metacharacter interpretation.
+- `auth_password` is tagged `json:"-"`; never serialized in responses, never
+  accepted from JSON request bodies.
+- Cross-host redirects (RM → NM 307) preserve the `Authorization` header
+  through `DoWithRedirect` only when the destination host is in
+  `url_allowlist`.
+- Per-Tool log files are created with mode `0600` and live under
+  `./data/logs/tools/`.
+
+## Development
+
+```bash
+go build ./...
+go test ./...
+go vet ./...
+```
+
+## More docs
+
+- [ARCHITECTURE.md](ARCHITECTURE.md) - Design and package layout
+- [docs/runbook.md](docs/runbook.md) - Deployment, upgrade, and troubleshooting
diff --git a/docs/runbook.md b/docs/runbook.md
new file mode 100644
index 0000000..70e4f17
--- /dev/null
+++ b/docs/runbook.md
@@ -0,0 +1,80 @@
+# spark-mcp-go 部署与排障手册
+
+## 部署步骤
+
+### 1. 构建
+
+```bash
+cd /opt/spark-mcp-go
+go build -o spark-mcp-go .
+```
+
+### 2. 配置
+
+```bash
+cp .env.example .env
+# 编辑 .env：ADMIN_TOKENS、AGENT_TOKEN、LISTEN_ADDR、DATA_DIR 等
+```
+
+### 3. 使用 systemd 运行
+
+参见 [docs/systemd/spark-mcp-go.service](systemd/spark-mcp-go.service)。
+
+### 4. 初始化示例 cluster
+
+```bash
+ADMIN_HOST=http://127.0.0.1:8080 ./scripts/seed.sh prod http://rm.example.com:8088 http://shs.example.com:18080 /opt/spark/bin/spark-submit
+```
+
+### 5. 升级流程
+
+1. 停止服务：`systemctl stop spark-mcp-go`
+2. 备份数据库：`cp ./data/spark-mcp.db ./data/spark-mcp.db.bak`
+3. 替换二进制：`cp spark-mcp-go /usr/local/bin/`
+4. 启动服务：`systemctl start spark-mcp-go`
+5. 验证：`curl -sS http://127.0.0.1:${LISTEN_ADDR:-:8080}/healthz`
+
+## 故障排查
+
+### 401 Unauthorized
+
+- 检查请求是否带 `Authorization: Bearer <token>`。
+- 区分 `ADMIN_TOKENS`（管理面）和 `AGENT_TOKEN`（MCP 面）。
+- `.env` 中 token 不能含前导/尾随空格。
+
+### 404 "Invalid session ID"
+
+- Streamable HTTP 需要先走 MCP `initialize` 握手，拿到 `Mcp-Session-Id` 后再调用 Tool。
+- 确保客户端保存并回传 response header 中的 session ID。
+
+### 422 cluster not found
+
+- `cluster_id` 拼写错误，或 cluster 未激活。
+- 先用 `list_clusters` 确认可用的 cluster ID。
+
+### "host not in allowlist"
+
+- 对应 cluster 的 `url_allowlist` 未包含目标主机。
+- 使用 `admin` 更新 cluster，把 RM/SHS 主机加入 allowlist。
+
+### `spark_submit` 返回 exit_code 7
+
+- `spark_submit_execute_bin` 路径错误，或 YARN 拒绝提交。
+- 查看返回的 `stderr_tail`；在服务器上直接运行相同命令验证。
+
+### `analyze_spark_log` 的 `findings` 为空
+
+- SHS metrics 端点 404 或返回格式不匹配。
+- 可 fallback 到 raw log + LLM 自行分析。
+
+### SQLite 锁冲突
+
+- `storage` 已开启 WAL 模式，但 SQLite 仍是 single-writer。
+- 多个 server 实例共用一个 `spark-mcp.db` 会产生锁竞争；请保持一实例一数据库。
+
+## 监控建议
+
+- **存活探测**：HTTP `GET /healthz`
+- **日志告警**：监控 `data/logs/spark-mcp.log` 中的 `ERROR` 级别行
+- **审计日志告警**：`data/logs/tools/` 下文件过大或增长过快时检查异常调用
+- **磁盘告警**：`data/` 目录包含 SQLite、上传文件、日志，需预留空间
diff --git a/docs/systemd/spark-mcp-go.service b/docs/systemd/spark-mcp-go.service
new file mode 100644
index 0000000..0d70598
--- /dev/null
+++ b/docs/systemd/spark-mcp-go.service
@@ -0,0 +1,22 @@
+[Unit]
+Description=Spark MCP Go Server
+After=network.target
+
+[Service]
+Type=simple
+User=spark-mcp
+Group=spark-mcp
+WorkingDirectory=/opt/spark-mcp-go
+EnvironmentFile=/opt/spark-mcp-go/.env
+ExecStart=/opt/spark-mcp-go/spark-mcp-go
+Restart=on-failure
+RestartSec=5
+
+# Security hardening
+NoNewPrivileges=true
+ProtectSystem=strict
+ProtectHome=true
+ReadWritePaths=/opt/spark-mcp-go/data
+
+[Install]
+WantedBy=multi-user.target
diff --git a/scripts/dev.sh b/scripts/dev.sh
new file mode 100755
index 0000000..eb455ff
--- /dev/null
+++ b/scripts/dev.sh
@@ -0,0 +1,45 @@
+#!/usr/bin/env bash
+# One-command build + start + health check for spark-mcp-go.
+set -euo pipefail
+
+cd "$(dirname "$0")/.."
+
+# Load .env if present
+if [ -f .env ]; then
+    set -a
+    # shellcheck source=/dev/null
+    . ./.env
+    set +a
+fi
+
+# Required tokens
+: "${ADMIN_TOKENS:?ADMIN_TOKENS is required}"
+: "${AGENT_TOKEN:?AGENT_TOKEN is required}"
+
+LISTEN_ADDR="${LISTEN_ADDR:-:8080}"
+PORT="${LISTEN_ADDR##*:}"
+PID_FILE="./data/spark-mcp.pid"
+
+echo ">> go build"
+go build -o ./spark-mcp-go .
+
+mkdir -p ./data/logs/tools
+
+echo ">> starting server on port ${PORT}"
+nohup ./spark-mcp-go > ./data/logs/spark-mcp.log 2>&1 &
+PID=$!
+echo "${PID}" > "${PID_FILE}"
+
+sleep 2
+
+echo ">> health check"
+if curl -sS "http://127.0.0.1:${PORT}/healthz"; then
+    echo
+    echo ">> running, PID=${PID}, log=./data/logs/spark-mcp.log"
+else
+    echo
+    echo ">> health check failed; see ./data/logs/spark-mcp.log"
+    kill "${PID}" 2>/dev/null || true
+    rm -f "${PID_FILE}"
+    exit 1
+fi
diff --git a/scripts/seed.sh b/scripts/seed.sh
new file mode 100755
index 0000000..906050a
--- /dev/null
+++ b/scripts/seed.sh
@@ -0,0 +1,33 @@
+#!/usr/bin/env bash
+# Admin POST to create a sample cluster.
+# Usage: ./scripts/seed.sh [cluster_id] [rm_url] [shs_url] [spark_submit_bin]
+set -euo pipefail
+
+cd "$(dirname "$0")/.."
+
+# Load .env if present
+if [ -f .env ]; then
+    set -a
+    # shellcheck source=/dev/null
+    . ./.env
+    set +a
+fi
+
+: "${ADMIN_TOKENS:?ADMIN_TOKENS is required}"
+
+HOST="${ADMIN_HOST:-http://127.0.0.1:8080}"
+TOKEN="${ADMIN_TOKENS%%,*}"
+AUTH="Authorization: Bearer ${TOKEN}"
+
+CLUSTER_ID="${1:-prod}"
+RM_URL="${2:-http://rm.example.com:8088}"
+SHS_URL="${3:-http://shs.example.com:18080}"
+SPARK_BIN="${4:-/opt/spark/bin/spark-submit}"
+
+echo ">> creating cluster '${CLUSTER_ID}' via ${HOST}/admin/clusters"
+
+curl -sS -X POST -H "${AUTH}" -H "Content-Type: application/json" \
+    -d "{\"id\":\"${CLUSTER_ID}\",\"name\":\"${CLUSTER_ID}\",\"rm_url\":\"${RM_URL}\",\"shs_url\":\"${SHS_URL}\",\"spark_submit_execute_bin\":\"${SPARK_BIN}\",\"is_active\":true,\"auth_type\":\"simple\",\"auth_username\":\"yarn\",\"rate_limit_per_min\":10,\"url_allowlist\":[\"${RM_URL#*://}\",\"${SHS_URL#*://}\"]}" \
+    "${HOST}/admin/clusters"
+
+echo