Space Memory
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Space Memorysearch memories for the decision on vector database"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Space Memory
Start with any agent. Continue with any other.
Space Memory is an open-source shared memory and workspace layer for AI agents. It persists attributable memories in an append-only event journal and lets multiple agent-specific keys access the same isolated Space through MCP or HTTP.
Early development preview. The current vertical slice proves durable memory, cross-agent reads, restart recovery, idempotency, optimistic concurrency, MCP tools, and the first living-brain web surface. It is not ready for production secrets.
What works now
Bearer keys stored as hashes
Separate agents sharing one Space
Append-only event journal
Durable memory projection
Keyword search without bundled AI
Idempotent create/update operations
Atomic optimistic version checks (
409on stale writes)Cursor-based event catch-up after restart
Per-credential rate limiting (sliding window,
429+Retry-After)Lifecycle CLI (
space-memory start|stop|status|restart) with PID + cmdline verificationAuthenticated web panel: Space-key login, validated against the same Bearer contract, with explicit invalid/offline/empty states and logout
MCP Streamable HTTP tools:
memory_remembermemory_searchproject_resumeevents_after
Locked dependency graph (
uv.lock)Animated Canvas dashboard with accessible textual fallback
Related MCP server: AgentBase
Development
Requires Python 3.11+.
python -m venv .venv
.venv/bin/pip install -e '.[dev]'
.venv/bin/pytestConfigure
Copy .env.example to .env and set a real pepper (generate once and keep it in
your secret store):
cp .env.example .env
# set SPACE_MEMORY_TOKEN_PEPPER (e.g. `openssl rand -hex 32`)The server and CLI auto-load .env from the current working directory. See
.env.example for every supported variable.
Run via the lifecycle CLI
.venv/bin/space-memory start # background, health-checked, PID tracked
.venv/bin/space-memory status
.venv/bin/space-memory stop # verifies the PID is a Space Memory process first
.venv/bin/space-memory restartstop never kills by port: it reads the PID file and checks /proc/<pid>/cmdline
before signalling, so an unrelated process that reused a stale PID is left alone.
Run manually (alternative)
mkdir -p data
export SPACE_MEMORY_TOKEN_PEPPER="replace-with-a-random-secret"
export SPACE_MEMORY_BOOTSTRAP_TOKEN="spm_local_replace_me"
export SPACE_MEMORY_SPACE_ID="my-space"
export SPACE_MEMORY_AGENT_ID="my-agent"
.venv/bin/uvicorn space_memory.main:app --host 127.0.0.1 --port 8787Open:
Panel:
http://127.0.0.1:8787MCP:
http://127.0.0.1:8787/mcp/Health:
http://127.0.0.1:8787/health
Never expose this development configuration directly to the internet. Production installation, PostgreSQL, TLS, admin onboarding, encrypted vault, backups, and signed installers are planned but not implemented yet.
Architecture and product record
License
MIT
This server cannot be deployed
Maintenance
Related MCP Connectors
Shared, governed long-term memory for AI agents across tools and sessions via MCP and REST.
- MemocoreOAuthai.memocore
Shared memory for all your AI agents, your whole team and every MCP client — save, search, recall.
Persistent personal memory for AI assistants — save, search, and recall across every MCP client.
Persistent AI memory shared across Claude, ChatGPT, coding agents, and compatible MCP clients.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceEnables persistent memory for AI agents, combining episodic and semantic memory with LLM reasoning, accessible via MCP.2MIT
- AlicenseNot gradedqualityDmaintenanceEnables AI agents to persistently store and semantically search shared knowledge via MCP tools.2MIT
- AlicenseNot gradedqualityCmaintenanceProvides persistent, cloud-based memory for AI agents, letting them remember preferences, skills, and decisions across sessions via an MCP-compatible interface.Apache 2.0
- AlicenseAqualityCmaintenanceEnables AI agents to share a portable, user-level memory layer through MCP, allowing them to store, search, update, link, and consolidate facts with optional full-text and vector retrieval.79MIT