feat: agent CLI transport, resilient MCP bridge, proactive Skill (0.2.0)

README.md

@@ -270,11 +270,29 @@ The integration is independently listed by
 
 ```sh
 sandbase connect [--client <name>]    # Authorize + configure
-sandbase doctor [--client <name>]     # Health check
+sandbase setup                        # After upgrading: refresh the local bridge + Skill, test connectivity
+sandbase doctor [--client <name>]     # Local checks plus one real round trip to SandBase
 sandbase unregister [--client <name>] # Remove configuration
 sandbase catalog --json               # List all supported clients
 ```
 
+### Agent Commands (shell transport, same capabilities as the MCP tools)
+
+Agents without MCP, or with a broken MCP session, use the same catalog from a shell. Credentials come from `connect` automatically (or `SANDBASE_API_KEY` / `sandbase keys add -k <key>`).
+
+```sh
+sandbase discover -q "twitter posts"                       # search; -t api|llm|multimodal|embedding|service
+sandbase inspect sandbase_twitter_web_search_timeline       # fields, price, ready-to-run command
+sandbase run sandbase_twitter_web_search_timeline -i '{"keyword":"AI agents"}'
+sandbase run <video_tool> -i '{"prompt":"..."}' -w -o result.json   # wait with backoff, save output to disk
+sandbase runs get -r <run_id> -w -o result.json            # resume a long run instead of re-running it
+sandbase runs list | sandbase balance
+```
+
+`-j` prints JSON; errors are `{"error":{"code","message","hint"}}` with the next step in `hint`. Exit codes: `0` success or still running, `1` failure (including a failed run), `2` usage.
+
+If `api.sandbase.ai` is unreachable, the CLI and the MCP bridge fall back to `www.sandbase.ai` automatically, and the bridge keeps the SandBase tools visible with an actionable error instead of dropping the server.
+
 ### MCP Tools (available to your agent after connecting)
 
 | Tool | Purpose |

README.zh-CN.md

@@ -204,11 +204,29 @@ npx -y https://github.com/sandbaseai/cli/releases/download/v0.1.17/sandbaseai-cl
 
 ```sh
 sandbase connect [--client <name>]    # 授权 + 配置
-sandbase doctor [--client <name>]     # 健康检查
+sandbase setup                        # 升级后：刷新本地 bridge 与 Skill，并测试连通性
+sandbase doctor [--client <name>]     # 本地检查 + 一次真实的 SandBase 往返请求
 sandbase unregister [--client <name>] # 移除配置
 sandbase catalog --json               # 列出所有支持的客户端
 ```
 
+### Agent 命令（Shell 通道，与 MCP 工具能力一致）
+
+没有 MCP、或 MCP 会话异常时，Agent 可以直接在 Shell 里使用同一套目录。凭据自动复用 `connect` 的浏览器登录结果（也可用 `SANDBASE_API_KEY` 或 `sandbase keys add -k <key>`）。
+
+```sh
+sandbase discover -q "推特搜索"                              # 搜索；-t api|llm|multimodal|embedding|service
+sandbase inspect sandbase_twitter_web_search_timeline       # 参数、价格、可直接运行的命令
+sandbase run sandbase_twitter_web_search_timeline -i '{"keyword":"AI agents"}'
+sandbase run <视频工具> -i '{"prompt":"..."}' -w -o result.json   # 带退避等待完成，结果写入文件
+sandbase runs get -r <run_id> -w -o result.json            # 续查长任务，不要重新运行
+sandbase runs list | sandbase balance
+```
+
+`-j` 输出 JSON；错误格式为 `{"error":{"code","message","hint"}}`，`hint` 给出下一步。退出码：`0` 成功或仍在运行，`1` 失败（含任务失败），`2` 参数错误。
+
+`api.sandbase.ai` 不可达时，CLI 与 MCP bridge 会自动切换到 `www.sandbase.ai`；bridge 会继续展示 SandBase 工具，并返回可执行的错误提示，而不是让客户端丢掉整个 server。
+
 ### MCP 工具（连接后你的 Agent 可以使用）
 
 | 工具 | 用途 |

package.json

@@ -1,8 +1,8 @@
 {
   "name": "@sandbaseai/cli",
-  "version": "0.1.17",
+  "version": "0.2.0",
   "mcpName": "io.github.sandbaseai/cli",
-  "description": "Connect 25 AI clients—including Claude Code, Codex, Cursor, Hermes, and OpenClaw—to 2,000+ models through a local MCP server",
+  "description": "Give 25 AI clients—Claude Code, Codex, Cursor, Hermes, OpenClaw and more—2,000+ models and APIs through one MCP server and an agent-friendly CLI",
   "type": "module",
   "bin": {
     "sandbase": "dist/cli.js"
@@ -20,7 +20,7 @@
   ],
   "scripts": {
     "build": "tsc -p tsconfig.json",
-    "test": "npm run build && tsc -p tsconfig.test.json && node --test dist-test/test/*.test.js",
+    "test": "npm run build && tsc -p tsconfig.test.json && node --import ./dist-test/test/setup.js --test dist-test/test/*.test.js",
     "lint": "tsc -p tsconfig.json --noEmit && tsc -p tsconfig.test.json --noEmit",
     "audit:package": "bash scripts/audit-package.sh",
     "prepack": "npm run build"

server.json

@@ -3,7 +3,7 @@
   "name": "io.github.sandbaseai/cli",
   "title": "SandBase",
   "description": "Connect MCP clients to 2,000+ AI models without managing provider API keys.",
-  "version": "0.1.17",
+  "version": "0.2.0",
   "repository": {
     "url": "https://github.com/sandbaseai/cli",
     "source": "github"
@@ -26,7 +26,7 @@
     {
       "registryType": "npm",
       "identifier": "@sandbaseai/cli",
-      "version": "0.1.17",
+      "version": "0.2.0",
       "packageArguments": [
         {
           "type": "positional",