AgentStack
Sign in Publish
MCP servers Skills Toolkits Why AgentStack Sell Docs
Sign in
MCP verified Apache-2.0 Self-run

Vistaclair

mcp-hpfrei-vistaclair · by hpfrei

Remote control room, agent inspector, and multi-provider model router for Claude Code

↗ Website ↗ Repository
— No reviews yet
0 installs
18 views
0.0% view→install

Install

$ agentstack add mcp-hpfrei-vistaclair

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • ✓ Prompt-injection patterns
  • ✓ Secret / credential exfiltration
  • ✓ Dangerous shell & filesystem operations
  • ✓ Untrusted network calls
  • ✓ Known-malicious package signatures

What it can access

  • ● Network access Used
  • ✓ Filesystem access No
  • ✓ Shell / process execution No
  • ✓ Environment & secrets No
  • ✓ Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Vistaclair? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

vistaclair

Agent inspector, control room, and transparent proxy for Claude Code.

[](https://www.npmjs.com/package/vistaclair) [](https://www.npmjs.com/package/vistaclair) [](https://opensource.org/licenses/Apache-2.0) [](https://nodejs.org)

npx vistaclair

I built this because I wanted to see what Claude Code actually sends to the LLM — the full requests, tool calls, hooks, subagent threads, token counts. The Inspector shows all of it on the wire, in real time, across sessions. From there it grew into a full control surface: multi-tab CLI sessions I can steer from a phone over a tunnel, proxy rules that rewrite or reroute requests to OpenAI/Gemini/DeepSeek/Ollama before they hit the API, and an integrated MCP server for custom tools. I use it daily.

https://github.com/user-attachments/assets/3d82bb12-87e1-4e20-aa98-52d4f3512062

Getting started

Requires: Node.js 18+ and Claude Code CLI (authenticated).

npm install -g vistaclair
vistaclair

Or clone:

git clone https://github.com/hpfrei/vistaclair.git
cd vistaclair && npm install && npm start

Open http://localhost:3457 and log in with the auth token printed to the console.

> [!TIP] > Use Claude through the proxy — run this in any terminal to route all traffic through vistaclair: > ``bash > ANTHROPIC_BASE_URL=http://localhost:3456 claude > ` > Works with interactive sessions, claude -p "prompt"`, and IDE integrations. Traffic appears in the Inspector.

> [!TIP] > Custom port: npm start -- 8080 or DASHBOARD_PORT=8080 npm start

Environment variables

| Variable | Default | Description | |---|---|---| | AUTH_TOKEN | auto-generated | Dashboard auth token | | PROXY_PORT | 3456 | API proxy port (localhost only) | | DASHBOARD_PORT | 3457 | Web dashboard port (all interfaces) | | ANTHROPIC_TARGET_URL | https://api.anthropic.com | Upstream API URL | | MAX_HISTORY | 200 | Max interactions kept in memory |

> [!WARNING] > The dashboard binds to 0.0.0.0 for remote access. It is protected by the auth token, but do not expose it to untrusted networks without TLS/VPN.

What it does

Inspector — Full request/response capture across all sessions. System prompts, message history, tool definitions, tool call inputs/outputs (JSON tree), thinking blocks, SSE event stream, cURL export. Token breakdown (input, output, cache read/create, reasoning) and cost tracking per turn, per group, per session. Subagent tracking with color-coded parallel swimlane view. Hook events nested under the turns that triggered them. All interactions saved to disk as JSON.

CLI — Multi-tab Claude Code terminal in the browser. Per-session model routing, directory spawning, session save/resume. Paste screenshots from clipboard or drag-drop files directly into a session — works over a tunnel from any device. AskUserQuestion appears inline — answer from any browser and Claude continues.

Provider routing — Route through Anthropic, OpenAI, Gemini, DeepSeek, Kimi, or Ollama. 40+ models pre-configured. The proxy translates Anthropic Messages API to the target format transparently.

Proxy rules — JavaScript middleware that intercepts every request. Swap models, strip tools, inject prompts, short-circuit responses. Describe in English, vistaclair generates the JS. Hot-reloaded, toggleable.

MCP tools — Define custom tools in the browser with typed parameters and a JS handler. Claude gets the capability instantly. Built-in: vista-AskUserQuestion (routes questions to dashboard UI), chat (sub-session delegation).

File manager — Browse, preview, and manage files on the remote machine. Monaco editor, multi-select, integrated shell.

Skills, agents, hooks — Create and edit .claude/ skills, agents, and hooks from the dashboard.

[External API](VISTACLAIR_CONNECT.md) — REST and WebSocket interface for programmatic access.

Architecture

Two servers, one process:

| Server | Port | Binding | Purpose | |---|---|---|---| | Proxy | :3456 | 127.0.0.1 | Intercepts Claude API calls for inspection and routing | | Dashboard | :3457 | 0.0.0.0 | Web UI, REST API, WebSocket (auth-protected) |

Every claude process gets a unique instance ID in its base URL (http://localhost:3456/i/{id}), so concurrent sessions are fully isolated.

License

Apache 2.0 — see [LICENSE](LICENSE). Copyright 2026 hpfreilabs.com

Everything above is free and open source. vistaclair Pro adds the Apps Platform as a separately licensed commercial add-on. Learn more →

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Write a review

Versions

  • v0.1.0 Imported from the upstream source.

More in AI & ML