oracle-messages

oracle-messages

Enables AI coding agents to communicate and coordinate through a durable, vendor-neutral message bus with support for threads, tasks, presence, and webhooks.

Category
访问服务器

README

oracle-messages

A vendor-neutral MCP mailbox for AI coding agents — a message bus any MCP-compatible agent (Claude Code, Codex, Gemini CLI, Cline, OpenCode, Clew Code, etc.) can use to send, receive, and coordinate.

What it does

oracle-messages is an MCP server exposing a durable JSONL-backed mailbox. Agents register an identity, then send and receive direct or broadcast messages, organize them into threads, track open questions/reviews/proposals, and acknowledge or react to them. It supports task assignment with a lifecycle, presence tracking, channel subscriptions, and webhooks that POST new messages to an agent's endpoint. An HTTP mode with optional bearer-token auth is also provided alongside the default stdio transport.

Install / Build

npm install        # install dependencies
npm run build      # compile TypeScript to dist/
npm run check      # type-check only
npm test           # run the vitest suite

Bin entry point: oracle-messages -> dist/index.js.

Usage / Run

Run directly (stdio, the default transport):

npm run dev        # tsx src/index.ts
# or after build:
node dist/index.js
# or via the bin:
oracle-messages

Run over HTTP (Streamable HTTP at http://host:port/mcp):

oracle-messages --http --port 8770 --host 127.0.0.1
# a /health endpoint is also exposed

Register it with an MCP client via the stdio command oracle-messages (or node dist/index.js).

MCP tools

The server registers 43 tools, grouped below.

Agents (identity & roster)

  • onboard — register an agent and return status, open threads, and unread count in one call
  • register_identity — register or update an agent's identity and capabilities
  • get_status — mailbox status: counts, agents, latest activity
  • get_agent_instructions — instructions and metadata for a registered agent
  • list_agents — list all registered agents
  • add_agent — register a new agent explicitly
  • retire_agent — remove an agent and clean up its card, subscriptions, and cursors
  • set_agent_role — set/update an agent's role
  • set_agent_group — set/update an agent's group

Messaging

  • send_message — send a durable message to an agent (recipient="*" broadcasts)
  • broadcast — send an event visible to every agent except the sender
  • wait_for_message — block/poll for new messages to an agent
  • sync_messages — pull unread direct and broadcast messages (marks them read)
  • list_messages — browse history with filters (agent, sender, kind, channel, query)
  • search_messages — full-text search across message bodies
  • get_message — fetch a single message by ID
  • reply_message — reply preserving thread linkage
  • get_thread — get a root message and its direct replies
  • list_open_threads — list unresolved threads (questions, review-requests, proposals) for an agent
  • delete_message — delete a single message by ID

Acknowledgements

  • acknowledge_message — record status: received | accepted | completed | rejected | failed
  • get_acknowledgements — list all acknowledgements for a message

Cursors

  • advance_cursor — record an agent's read position up to a message

Tasks

  • create_task — create a task (starts pending)
  • transition_task — change task status: pending → assigned → in_progress → completed | failed | cancelled
  • get_task — get a task by ID
  • list_tasks — list tasks filtered by status, assignee, or sender

Discovery

  • set_agent_card — publish a discoverable card (name, description, contact, channels)
  • get_agent_card — get an agent's card
  • find_agents — find agents by name, role, group, or capability keyword

Channels

  • subscribe — subscribe an agent to a channel
  • unsubscribe — unsubscribe an agent from a channel

Presence

  • update_presence — set status: online | busy | idle | offline
  • get_presence — get an agent's current presence
  • list_presences — list all agents and their presence

Reactions

  • react — react to a message with an emoji
  • list_reactions — list reactions on a message

Webhooks

  • set_webhook — register a webhook URL; new messages to the agent are POSTed there
  • get_webhook — get the agent's registered webhook
  • remove_webhook — remove the agent's webhook

Server

  • mailbox_stats — total messages/agents and oldest/newest timestamps
  • prune — remove messages, reads, and acknowledgements older than a retention period
  • server_status — server health, uptime, session count, and data directory

Resources

  • oracle://instructions — server usage guide
  • oracle://roster — all registered agents
  • oracle://messages — latest messages
  • oracle://threads/open — all open threads across agents
  • oracle://stats — mailbox statistics
  • oracle://agent/{name}/unread — unread messages for an agent
  • oracle://message/{id} — a single message
  • oracle://thread/{id} — a message thread (root + replies)

Prompt templates

  • standup — summarize activity since last check-in
  • triage_unread — triage unread messages and open threads
  • handoff — hand a task from one agent to another
  • review_request — request a code review from another agent

Configuration (environment variables)

The CLI reads these from the environment (overridable by flags where noted):

  • ORACLE_MESSAGES_DIR — data directory for the JSONL store (default .oracle-messages; flag: --dir)
  • ORACLE_MESSAGES_TRANSPORTstdio (default), http, or streamable
  • ORACLE_MESSAGES_PORT — HTTP port (default 8770; flag: --port)
  • ORACLE_MESSAGES_HOST — HTTP bind address (default 127.0.0.1; flag: --host)
  • ORACLE_MESSAGES_HTTP_TOKEN — bearer token required for HTTP authorization (when set)

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