mcp-blackboard

mcp-blackboard is a server designed to facilitate collaboration among AI agents by providing a shared 'blackboard' where context and results can be stored and retrieved. It offers a simple HTTP/SSE interface, making it easy to integrate into any MCP-compatible workflow. This allows various agents such as planners, researchers, extractors, analyzers, writers, editors, and evaluators to work together without the need for separate persistence mechanisms. The server supports storing documents, embeddings, structured objects, and more, ensuring a unified memory source for agent context. Built on fsspec, it offers filesystem abstraction with optional drivers for various storage solutions like S3, Azure Blob, and more. Real-time updates are facilitated through Server-Sent Events (SSE), and a housekeeping scheduler is available for automatic maintenance tasks. The server is container-ready, with deterministic builds and a slim Docker image.

Features

• {'name': 'Unified memory', 'description': 'Provides a single source of truth for agent context, eliminating the need for ad-hoc scratch files or transient Redis keys.'}
• {'name': 'Filesystem abstraction', 'description': 'Built on fsspec with optional drivers for S3, Azure Blob, GCS, ABFS, SFTP, SMB, and more.'}
• {'name': 'Real-time updates', 'description': 'Server-Sent Events (SSE) stream context changes to connected agents instantly.'}
• {'name': 'House-keeping scheduler', 'description': 'Pluggable cron jobs automatically prune expired keys and refresh embeddings.'}
• {'name': 'Container-ready', 'description': 'Deterministic builds via uv lockfile; the slim Docker image is less than 90 MB.'}

MCP Tools

• save_plan: Save a plan to the shared state.
• mark_plan_as_completed: Mark a plan step as completed in the shared state.
• save_result: Save a result to the shared state.
• save_context_description: Write a context description to the shared state.
• get_blackboard: Fetch a blackboard entry for a plan.
• get_plan: Fetch a plan from the shared state.
• get_result: Fetch a result from the shared state.
• get_context: Read and convert media content to Markdown format.
• remove_stale_files: Remove files older than the specified age from the cache directory.

Usage with Different Platforms

Local dev environment

bash
git clone https://github.com/your-org/mcp-blackboard.git
cd mcp-blackboard

# Create an isolated env & install locked deps
uv venv
uv sync

# Copy the sample env and fill in credentials
cp samples/env-sample.txt .env

# Run the server (FastAPI)
uv run src/main.py

Docker Compose

bash
docker compose up -d

Compose starts:
- **mcp-blackboard** – FastAPI+SSE service
- **redis** – in-memory store for keys, scores, embeddings