BrainHub
Provides a tool to send Matrix messages via the BrainBridge daemon, enabling agents to communicate through the Matrix protocol.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
BrainHub — Unified Web Frontend + File Storage + MCP Gateway + Ops Agent
The self-use main entry of the Brain product family: a single FastAPI process aggregating all panels (knowledge base / search / file storage / kanban / memory / agent status / ops logs), exposing a single MCP gateway externally, with a built-in ops agent that automatically maintains the knowledge base and memory.
Introduction
BrainHub is the entry point for the Brain product family, consolidating all capabilities into one operable interface and one set of APIs:
Web Panel: Jinja2 + HTMX zero-frontend build, knowledge base browsing / semantic search / file storage CRUD / kanban task state machine / memory / agent status / ops logs
JSON API v1:
/api/v1/*for the desktop React frontendMCP Gateway: aggregates 12 tools total from BrainHub's own tools + BrainMem retrieval tools; any agent only connects to this one entry
Ops Agent: scheduled archiving, index rebuild, memory extraction, health checks
In the self-use phase, a single brainhub start brings up the entire main process: imports brainmem as the retrieval backend, spawns the brain-bridge subprocess, appearing as one software externally.
Related MCP server: stac2026
Features
🖥️ Unified Panel — 7 types of pages aggregated in one service, HTMX partial refresh + Alpine lightweight interactions, no frontend build chain needed
🔌 Single MCP Entry —
write_note/read_file/list_files/ project state machine /health_check/send_matrix_msg+ forwarding BrainMem retrieval tools🤖 Ops Agent — APScheduler scheduled: archiving / index rebuild / memory extraction / health checks, no external orchestration needed
📁 File Storage — BRAIN_ROOT file browsing, upload/download, lazy thumbnail generation, out-of-bounds and sensitive path interception
📋 Kanban — project/task state machine (todo → doing → blocked → done, strict transitions)
🔗 Collaboration Integration — forwards
send_matrix_msgto the BrainBridge daemon via named pipe, frontend receives Matrix messages and agent status in real-time via WS💻 Desktop Same-Origin Mount — mounts brainhub-desktop build artifacts at
/app, same-origin without CORS
Architecture
Directory | Responsibility |
| FastAPI app + Jinja2 templates + WS; |
| File CRUD + thumbnails, |
| Project/task state machine models |
| Ops agent: memory extraction, OpsAgent interface, APScheduler scheduling |
| Named pipe client (communicates with BrainBridge, envelope aligned with protocol constitution §3) |
| fastmcp gateway: own tools + forwarding BrainMem tools |
|
|
Quick Start
# 安装(Python 3.10+,uv 管理;依赖 BrainMem editable 安装)
uv sync
uv pip install -e ../brainmem
# 启动(拉起 Web + MCP 网关 + brain-bridge 子进程)
uv run brainhub start
# 运维子命令
uv run brainhub ops archive
uv run brainhub ops reindex
uv run brainhub ops healthDocumentation Index
PLAN.md — implementation plan and current progress (Phase 1-3 acceptance, panel, ops Cron, MCP tools, dependencies)
docs/protocol-constitution.md — protocol constitution: MCP tool signatures / memory.db schema / Matrix envelope
../brainmem/PRODUCT_FAMILY_PLAN.md — product family global plan (cross-product boundaries, HX470 migration, risk trade-offs)
Related Projects
Brain product family: BrainMem (memory engine) · BrainHub (this repository) · BrainBridge (collaboration gateway)
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
MCP server unifying ERPs, CRMs, APIs and knowledge base for Claude, ChatGPT and Gemini.
One MCP server for the services your AI needs: memory, files, Microsoft 365, kanban and more.
MCP gateway and cloud hosting for MCP servers, agents and apps like RAG, managed from your agent.
Agent-native notes, tasks, dev-docs, vaults, sync & handoffs. MCP + OpenAPI dual surface.
Related MCP Servers
- FlicenseNot gradedqualityNot gradedmaintenanceProvides a unified MCP interface that proxies requests to multiple backend servers including memory/knowledge graph and vector database services. Enables seamless access to distributed MCP tools through a single endpoint with automatic routing, health monitoring, and retry logic.-
- FlicenseNot gradedqualityDmaintenancePersonal AI infrastructure gateway orchestrating memory, workflow automation, knowledge retrieval, and skill discovery as a unified MCP stack.-
- AlicenseNot gradedqualityBmaintenanceMCP server providing persistent memory, semantic search, versioned storage, webhook fanout, and unified LLM routing for AI agents via FastAPI runtime with multiple backend options.35Apache 2.0
- AlicenseNot gradedqualityDmaintenanceA FastAPI + FastMCP shared-memory gateway in front of Honcho v3 that provides a single shared brain for multiple AI agents to read from and write to, enabling context recall and change logging through bearer-authenticated endpoints.2MIT