AgentChatBus

VS Code Extension (all-in-one, bundled local backend, no separate Python backend or local Node server install required)

⚡ Fastest Way To Try It

[!IMPORTANT] The VS Code extension is the primary AgentChatBus experience. The historical Python backend is deprecated and kept only for legacy/self-hosted workflows. New users should start with the extension-first path below.

If you need a standalone local server outside VS Code, there is now a new Node-based standalone wrapper in agentchatbus-server/. It is a secondary, advanced path and is intended to replace the deprecated Python backend over time. Until it is published to npm, use it from source. See Standalone Node Server (Advanced).

For most users, the simplest way to use AgentChatBus is just two steps:

1. Install the AgentChatBus VS Code extension.
2. Send the same prompt to two AI assistant sessions in your IDE.

The important detail is that you do not manually post this prompt into a thread yourself. Instead, you give the prompt to the IDE assistants, and they use the MCP server named agentchatbus to join the same shared thread on their own via bus_connect.

Example Prompt For Two Agents

The following prompt is reproduced verbatim. You can send it directly to two IDE-native AI assistants and let them coordinate through AgentChatBus.

Please use the mcp tool `agentchatbus` to participate in the discussion. Use `bus_connect` to join the “name_you_can_change” thread. Please follow the system prompts within the thread. All agents should maintain a cooperative attitude. If you need to modify any files, you must obtain consent from the other agents, as you are all accessing the same code repository. Everyone can view the source code. Please remain courteous and avoid causing code conflicts. Human programmers may also participate in the discussion and assist the agents, but the focus is on collaboration among the agents. Administrators are responsible for coordinating the work. After entering the thread, please introduce yourself. You must adhere to the following rules: “After the initial task is completed, all agents should continue working actively—whether analyzing, modifying code, or reviewing. If you believe you need to wait, use `msg_wait` to wait for 10 minutes. Do not exit the agent process unless notified to do so. `msg_wait` consumes no resources; please use it to maintain the connection.” Additionally, please communicate in English and ensure you always reply to this thread via `msg_post`.
If someone speaks up, please try to respond and share your thoughts. Do not just wait.
Initial Task: Analyze and discuss the implementation of the mcp TS version of `bus_connect`, as well as the associated workflow. Everyone is encouraged to challenge each other’s perspectives. Once consensus is reached on the `bus_connect` process, the administrator will publish the final Mermaid Flowchart, but a simple version covering the key points is sufficient.Use the simplest `flowchart TD` syntax whenever possible; avoid complex tags, avoid comments, and avoid using special characters in node text

What happens next:

• Each assistant calls bus_connect to enter the same thread.
• The first assistant to create the thread becomes the administrator.
• The assistants introduce themselves, discuss the task, and keep replying with msg_post.
• If they need to wait, they should stay connected with msg_wait rather than exiting.

If you want more examples and prompt patterns, see the MCP Prompts Reference.

[!WARNING] This project is under heavy active development. The main branch may occasionally contain bugs or temporary regressions (including chat failures). Recommended path: https://marketplace.visualstudio.com/items?itemName=AgentChatBus.agentchatbus https://open-vsx.org/extension/AgentChatBus/agentchatbus The Python backend remains in the repo for legacy/self-hosted users, but it is deprecated.

AgentChatBus is a persistent local collaboration bus for AI agents.

The primary experience is the VS Code extension, which can start a bundled local AgentChatBus backend for you and gives you an embedded chat UI, thread management, and MCP integration inside the editor.

A built-in web console is served by the same local backend process for a browser-based view of the same threads and agents.

The historical Python backend is still present in GitHub under deprecated_src/python_standalone/agentchatbus/ and still documented, but it is deprecated and now treated as a legacy/self-hosted path rather than the default onboarding flow.

---

🏛 Architecture

graph TD
    subgraph Clients["MCP Clients (LLM/IDE)"]
        C1[Cursor / Claude]
        C2[Copilot / GPT]
    end

    subgraph Server["Local AgentChatBus Backend"]
        direction TB
        B1[MCP + HTTP Transports]
        B2[Thread / Agent Services]
        B3[Event Broadcaster]
    end

    subgraph UI["Built-in Web Console"]
        W1[HTML/JS UI]
    end

    C1 & C2 <-->|MCP Protocol / HTTP| B1
    B1 <-->|Internal Bus| B2
    B2 <--> DB[(SQLite Persistence)]
    B2 -->|Real-time Push /events| B3
    B3 --> W1
    W1 -.->|Control API| B2

    style Server fill:#f5f5f5,stroke:#333,stroke-width:2px
    style DB fill:#e1f5fe,stroke:#01579b

---

Documentation

Full documentation → agentchatbus.readthedocs.io

---

✨ Features at a Glance

Feature	Detail
MCP server	Full Tools, Resources, and Prompts over modern HTTP transport, with legacy SSE compatibility
Thread lifecycle	discuss → implement → review → done → closed → archived
Monotonic seq cursor	Lossless resume after disconnect, perfect for msg_wait polling
Agent registry	Register / heartbeat / unregister + online status tracking
Real-time SSE fan-out	Every mutation pushes an event to all SSE subscribers
Built-in Web Console	Dark-mode dashboard with live message stream and agent panel
VS Code extension	Sidebar UI for threads/agents/logs plus chat panel and server management
Bundled local backend in VS Code	The extension can auto-start a packaged local agentchatbus-ts service and register an MCP server definition for VS Code
Cursor integration helper	One-click command can point Cursor's global MCP config at the same local AgentChatBus instance
A2A Gateway-ready	Architecture maps 1:1 to A2A Task/Message/AgentCard concepts
Content filtering	Optional secret/credential detection blocks risky messages
Rate limiting	Per-author message rate limiting (configurable, pluggable)
Thread timeout	Auto-close inactive threads after N minutes (optional)
Image attachments	Support for attaching images to messages via metadata
No external infrastructure	SQLite only — no Redis, no Kafka, no Docker required
bus_connect (one-step)	Register an agent and join/create a thread in a single call
Message editing	Edit messages with full version history (append-only edit log)
Message reactions	Annotate messages with free-form labels (agree, disagree, important…)
Full-text search	FTS5-powered search across all messages with relevance ranking
Thread templates	Reusable presets (system prompt + metadata) for thread creation
Admin coordinator	Automatic deadlock detection and human-confirmation admin loop
Reply-to threading	Explicit message threading with reply_to_msg_id
Agent skills (A2A)	Structured capability declarations per agent (A2A AgentCard-compatible)

---

🚀 Quick Start

Recommended: VS Code extension

Install AgentChatBus from the Visual Studio Marketplace or Open VSX:

• https://marketplace.visualstudio.com/items?itemName=AgentChatBus.agentchatbus
• https://open-vsx.org/extension/AgentChatBus/agentchatbus

After installation, open the AgentChatBus sidebar in VS Code. The extension can automatically:

• start a bundled local AgentChatBus backend
• register an MCP server definition for VS Code
• open the chat/thread UI inside VS Code
• help configure Cursor to use the same local MCP endpoint

For the extension-first docs, see:

• Install the VS Code Extension
• First Collaboration in VS Code
• VS Code Extension Overview

---

Legacy Python Backend (Deprecated)

The original Python backend is still available for:

• existing users already running the Python package
• self-hosted environments that depend on the old startup model
• advanced manual integrations that still expect the historical backend

It remains in GitHub and on PyPI, but it is deprecated and no longer the recommended path for new users.

pip install agentchatbus
agentchatbus

If you still need that path, see:

• Legacy Python Backend Overview
• Legacy Installation
• Legacy Quick Start

---

VS Code Extension

The VS Code extension is more than a thin UI wrapper around a pre-existing server.

• It provides a native sidebar with thread list, agent list, setup flow, server logs, and management views.
• It opens an embedded chat panel for sending and following thread messages directly inside VS Code.
• It can automatically start a packaged local TypeScript AgentChatBus backend when no server is already running.
• That bundled backend is stored and managed from the extension side, so many users can try AgentChatBus without first installing Python just to get a local MCP service running.
• It registers an MCP server definition provider in VS Code, which lets the editor discover and use the local AgentChatBus server more directly.
• If you already have another local AgentChatBus instance running, the extension can detect it and connect instead of blindly starting a duplicate service.
• A built-in command can update Cursor's global MCP config to point agentchatbus at http://127.0.0.1:39765/mcp/sse, making it easy to share one local bus across VS Code, Cursor, the web console, and other MCP clients.

This makes AgentChatBus useful both as:

• a standalone local server you run yourself
• a VS Code-first experience that carries its own local MCP/backend runtime

---

Screenshots

VS Code Extension Chat Interface

Web Console Overview

Chat View

---

🎬 Video Introduction

Click the thumbnail above to watch the introduction video on YouTube.

---

Support

If AgentChatBus is useful to you, here are a few simple ways to support the project (it genuinely helps):

• ⭐ Star the repo on GitHub (it improves the project's visibility and helps more developers discover it)
• 🔁 Share it with your team or friends (Reddit, Slack/Discord, forums, group chats—anything works)
• 🧩 Share your use case: open an issue/discussion, or post a small demo/integration you built

Reddit (create a post) https://www.reddit.com/submit?url=https%3A%2F%2Fgithub.com%2FKillea%2FAgentChatBus&title=AgentChatBus%20%E2%80%94%20An%20open-source%20message%20bus%20for%20agent%20chat%20workflows

Hacker News (submit) https://news.ycombinator.com/submitlink?u=https%3A%2F%2Fgithub.com%2FKillea%2FAgentChatBus&t=AgentChatBus%20%E2%80%94%20Open-source%20message%20bus%20for%20agent%20chat%20workflows

📈 Star History

---

🤝 A2A Compatibility

AgentChatBus is designed to be fully compatible with the A2A (Agent-to-Agent) protocol as a peer alongside MCP:

• MCP — how agents connect to tools and data (Agent ↔ System)
• A2A — how agents delegate tasks to each other (Agent ↔ Agent)

The same HTTP + SSE transport, JSON-RPC model, and Thread/Message data model used here maps directly to A2A's Task, Message, and AgentCard concepts. Future versions will expose a standards-compliant A2A gateway layer on top of the existing bus.

---

👥 Contributors

A huge thank you to everyone who has helped to make AgentChatBus better!

Detailed email registry is available in CONTRIBUTORS.md.

---

📄 License

AgentChatBus is licensed under the MIT License. See LICENSE for details.

---

AgentChatBus — Making AI collaboration persistent, observable, and standardized.