It is an HTTP + MCP personal memory server that manages a graph knowledge base, technical observations, and conversation transcripts.
Personal graph memory: initialize/seed a database, create custom node/edge types, upsert entities, and link entities with typed directed edges.
Link & disambiguate: find candidate entities before creating them, merge duplicate nodes, and preserve provenance.
Capture reflections & attachments: store raw text as documents linked to entities, and attach files/images with SHA-based deduplication.
Explore & retrieve: get node neighborhoods, full node contents, provenance documents, and keyword (FTS5) search across names, summaries, and document text.
Visualize: fetch graph-ready subgraphs with color/icon metadata for rendering.
Technical memory (Engram-style): save, search, get, update, and list open observations/tasks with project/status/type metadata.
Conversation audit: import raw AI console transcripts, search them, and retrieve full conversation contents.
Click on "Install 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., "@hippocampus store this article about MCP and link it to my research node"
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.
Personal graph memory with an HTTP + MCP surface, a replacement for Obsidian/Engram.
Stack: uv · Python 3.11+ · SQLModel + SQLite (FTS5) · FastAPI + FastMCP · Typer · Rich.
uv sync # OBLIGATORIO
uv run hippocampus init <db-path> # bootstrap (crea BD + siembra)
uv run hippocampus serve # HTTP+MCP en 127.0.0.1:27130Subcommands: init, serve, mcp (stdio), graph (interactive HTML map, --tui for the table view), backup (SQL dump).
Related MCP server: 2BToRePensieve
Clean Architecture by layers (dependencies → inward):
api/ — HTTP/MCP transport (zero logic).
schemas/ — API contract (Pydantic, D26).
core/ — Domain (search-and-link, graph, technique).
db/, models/ — Persistence (SQLModel, FTS5, seed).
Two isolated spaces by construction (no space column):
Personal (graph): nodes → edges → documents + attachments. Replaces Obsidian.
Technical (Engram-style): observations with status/type. Replaces Engram.
See docs/sdd-hippocampus.md (complete SDD) and docs/todo-hippocampus.md (checklist) for details.
docs/architecture.md — principles and structure.
docs/data-model.md — ER schema (8 tables + FTS).
docs/core-services.md — domain services (search-and-link).
docs/http-mcp-api.md — transport (15 tools).
docs/e2e-flows.md — real verified flows.
docs/startup-and-init.md — startup + NO_INIT.
docs/untime-launchd.md — session autostart.
docs/workspace-rewiring.md — integration with psychologist.
docs/specs/sdd-hippocampus.md — detailed design (decisions D1–D26, requirements R1–R13).
docs/specs/todo-hippocampus.md — implementation checklist (Phase 0 complete).
docs/specs/sdd-obsidian-migration.md — deterministic Obsidian→hippocampus migrator.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Personal knowledge graph as an AI memory layer over MCP - read, save, and link your memories.
Persistent, portable memory for AI assistants — your private memory graph, from any MCP client.
Private-by-default, local-first memory/context/task orchestrator for MCP apps and agents.
Token-efficient MCP memory for Markdown vaults. Tiered search, GraphRAG, AI memories.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/kaidewu01/hippocampus'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
