Tesseron

Tesseron

Enables MCP-compatible AI agents to execute typed live-app actions over WebSocket, allowing agents to directly interact with real application state without browser automation or scraping.

Category
访问服务器

README

<div align="center">

<picture> <source media="(prefers-color-scheme: dark)" srcset="./assets/logo/tesseron-smallcaps-dark.png"> <img src="./assets/logo/tesseron-smallcaps-light.png" alt="Tesseron" width="520"> </picture>

Typed live-app actions for MCP-compatible AI agents, over WebSocket.

<p> <a href="https://github.com/BrainBlend-AI/tesseron/stargazers"><img alt="GitHub stars" src="https://img.shields.io/github/stars/BrainBlend-AI/tesseron?style=flat-square&color=f59e0b&logo=github&labelColor=0b1220"></a> <a href="./LICENSE"><img alt="License: BUSL-1.1" src="https://img.shields.io/badge/License-BUSL--1.1-f59e0b?style=flat-square&labelColor=0b1220"></a> <a href="https://discord.gg/J3W9b5AZJR"><img alt="Discord" src="https://img.shields.io/badge/chat-on%20discord-7289DA?logo=discord&style=flat-square&labelColor=0b1220"></a> <img alt="Protocol 1.0.0" src="https://img.shields.io/badge/Protocol-1.0.0-f59e0b?style=flat-square&labelColor=0b1220"> <img alt="TypeScript" src="https://img.shields.io/badge/TypeScript-5.7-3178c6?style=flat-square&logo=typescript&logoColor=white&labelColor=0b1220"> <img alt="Node 20+" src="https://img.shields.io/badge/Node-%E2%89%A5%2020-339933?style=flat-square&logo=node.js&logoColor=white&labelColor=0b1220"> <img alt="Tests" src="https://img.shields.io/badge/Tests-65%20passing-22c55e?style=flat-square&labelColor=0b1220"> </p>

<p> <a href="https://brainblend-ai.github.io/tesseron/"><b>Docs</b></a>  ·  <a href="./examples"><b>Examples</b></a>  ·  <a href="#quick-install-claude-code"><b>Install</b></a>  ·  <a href="#packages"><b>Packages</b></a>  ·  <a href="https://discord.gg/J3W9b5AZJR"><b>Discord</b></a>  ·  <a href="https://github.com/BrainBlend-AI/tesseron/discussions"><b>Discussions</b></a> </p>

<a href="https://brainblendai.com/"><img src="./assets/brainblend-ai/logo.svg" width="22" height="18" align="center" alt="BrainBlend AI"></a> <sub>Built by <a href="https://brainblendai.com/"><b>BrainBlend AI</b></a></sub>

</div>


NEW: Join our community on Discord at discord.gg/J3W9b5AZJR — protocol questions, feedback, and SDK-contribution chat all welcome.

Live applications (browser tabs, Electron/Tauri desktop apps, Node daemons, CLIs) declare actions with a Zod-style builder; agents (Claude Code, Claude Desktop, Cursor, Copilot, Codex, Cline, ...) call them as MCP tools. Your real handler runs against your real state. No browser automation, no scraping, no Playwright.

<p align="center"> <img src="./assets/diagrams/pieces-fit-together.png" alt="USER prompts the agent; YOUR APP (browser or Node, using @tesseron/web or /server) connects over WebSocket to the MCP GATEWAY (@tesseron/mcp on :7475); the MCP GATEWAY bridges to the MCP CLIENT (Claude Code, Desktop, Cursor) over MCP stdio." width="900"> </p>

Why Tesseron

  • Typed actions, not scraped DOMs. Declare with Zod or any Standard Schema validator; the handler is a plain function against your real state.
  • Framework-agnostic. Same API for vanilla TS, React, Svelte, Vue, and Node. Pick your stack.
  • MCP-native. Every action, resource, and capability maps to a standard MCP primitive. Users pick their agent.
  • Click-to-connect. Six-character claim code handshake. No API keys, no OAuth dance, no per-client configuration.
  • First-class capabilities. ctx.confirm for yes/no, ctx.elicit for schema-validated prompts, ctx.sample for agent LLM calls, ctx.progress for streaming updates, subscribable resources for live reads.
  • Bundled delivery. The MCP gateway ships inside a Claude Code plugin — one install command and you're done.

Quick install (Claude Code)

/plugin marketplace add BrainBlend-AI/tesseron
/plugin install tesseron@tesseron

That installs the tesseron Claude Code plugin, which spawns the MCP gateway automatically and registers it as an MCP server. Then drop @tesseron/web, @tesseron/server, or @tesseron/react into your app, declare actions, and let Claude drive your real UI.

import { tesseron } from '@tesseron/web';
import { z } from 'zod';

tesseron.app({ id: 'todo_app', name: 'Todo App' });

tesseron
  .action('addTodo')
  .input(z.object({ text: z.string().min(1) }))
  .handler(({ text }) => {
    state.todos.push({ id: newId(), text, done: false });
    render();
    return { ok: true };
  });

await tesseron.connect();

For other MCP clients (Claude Desktop, Cursor, Codex, VS Code Copilot, ...) see the one-time setup in examples/README.md.

See examples/ for working apps in vanilla TS, React, Svelte, Vue, Express, and plain Node.

Packages

Package Purpose
@tesseron/core Protocol types, action builder. Zero runtime deps beyond Standard Schema.
@tesseron/web Browser SDK.
@tesseron/server Node SDK.
@tesseron/mcp MCP gateway server (CLI; bundled into the plugin).
@tesseron/react React hooks adapter.
@tesseron/devtools In-browser debug UI served by the MCP gateway (private stub, not yet published).
create-tesseron npm create tesseron@latest scaffolder (private stub, not yet published).

The Claude Code plugin lives at plugin/, exposed via the marketplace manifest at .claude-plugin/marketplace.json.

Client capability support

Tesseron's action context gives handlers four capabilities beyond plain tool invocation, each backed by an MCP primitive. Whether a given call actually fires depends on what the user's MCP client advertises:

SDK surface MCP primitive
tool(...) (action invocation) tools
resource(...) (live reads, subscriptions) resources (+ resources.subscribe)
ctx.sample(...) sampling
ctx.confirm(...) / ctx.elicit(...) elicitation
ctx.progress(...) notifications/progress (client must pass _meta.progressToken on tools/call)

For the authoritative, continuously-updated list of which client supports which primitive, see the official MCP client compatibility matrix — filter by Sampling or Elicitation to see how narrow the field still is. A few points worth knowing before you pick a capability:

  • Tools are universal — every MCP client can invoke your actions.
  • Sampling is the rarest. Claude Code, Claude Desktop, and Claude.ai do not expose it; today's support is concentrated in VS Code + GitHub Copilot, goose, and fast-agent.
  • Elicitation (MCP 2025-06) landed in Claude Code (2.1.76, March 2026), Cursor, Codex, VS Code Copilot, goose, and fast-agent, but not Claude Desktop, Claude.ai, ChatGPT, Windsurf, or Zed.
  • When a capability is missing, Tesseron raises a typed error (SamplingNotAvailableError, ElicitationNotAvailableError) or collapses to the safe default (ctx.confirm returns false), so handlers can branch explicitly rather than silently misbehaving.

Status

v1.0 shipped April 2026. The protocol is stable at 1.0.0 and intentionally kept small: bidirectional JSON-RPC 2.0 over WebSocket, dynamic MCP tool registration, click-to-connect handshake, streaming progress, cancellation, sampling, confirmation, schema-validated elicitation, subscribable resources.

Published to npm (all at v1.0.1): @tesseron/core, @tesseron/web, @tesseron/server, @tesseron/react, @tesseron/mcp. The JS/TS SDKs are the reference implementation; the protocol spec is CC BY 4.0 so anyone can write a compatible client or server in any language.

On the roadmap: Svelte/Vue adapters, the devtools UI, a Streamable HTTP transport, a Python SDK, and bindings for desktop-native runtimes (Rust for Tauri, etc.).

Development

pnpm install
pnpm typecheck
pnpm test                                    # 65 tests across core + mcp
pnpm --filter @tesseron/mcp build:plugin     # rebuild plugin/server/index.cjs after gateway changes

Contributing

Bug reports, protocol refinements, new framework adapters, and improvements to the reference implementation are welcome.

Star history

<a href="https://star-history.com/#BrainBlend-AI/tesseron&Date"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/svg?repos=BrainBlend-AI/tesseron&type=Date&theme=dark"> <img alt="Star History Chart" src="https://api.star-history.com/svg?repos=BrainBlend-AI/tesseron&type=Date" width="720"> </picture> </a>

License

Reference implementationBusiness Source License 1.1 (source-available). You may embed Tesseron in your own applications, use it internally, fork it, and redistribute it freely. You may not offer Tesseron or a substantial portion of it as a hosted or managed service to third parties. Each release auto-converts to Apache-2.0 four years after publication.

Protocol specificationCC BY 4.0. A compatible implementation in any language, for any purpose including commercial, is explicitly encouraged.

Contributions are welcome under the Developer Certificate of Origin — every commit must be Signed-off-by.


<p align="center"> <a href="https://brainblendai.com/"><img src="./assets/brainblend-ai/logo.svg" width="32" height="27" alt="BrainBlend AI"></a> </p> <p align="center"> Built and maintained by <a href="https://brainblendai.com/"><b>BrainBlend AI</b></a>.<br> © 2026 Kenny Vaneetvelde. </p>

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选