Aldus Palace MCP
Aldus Palace is a memory MCP server that sorts sentences into commitments and decisions with evidence in your own SQLite file.
- 95 ★ on the repository
- Last commit
- Apache-2.0
Minimal access
Why this level
- Personal notes and preferences are kept in a plain local file without encryption.
Install
- Works in
Manual install
npm install -g @aldus-palace/mcp
claude mcp add aldus-palace -- node "$(npm root -g)/@aldus-palace/mcp/dist/index.js"Two README commands: package install and registration in Claude Code.
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- Personal notes and preferences are kept in a plain local file without encryption.
This is third-party code. Review the repository files before installing.
What it does
Aldus Palace takes free text and sorts it into typed objects: dated commitments, decisions, thoughts and memory candidates. Each object keeps the source sentence, a confidence value and a note on whether the user stated it or the system inferred it. Data lives in one local SQLite file. Claude, Cursor and scripts read the same file through MCP, an HTTP API, a library or a typed client. Repeated sentences are detected and point to the first record.
Who it is for. Developers of AI products and people who want assistant memory in their own file rather than inside a closed service.
Good fit when
- User tasks and preferences must carry over between Claude, Cursor and a terminal.
- An assistant needs memory where every fact shows its source and confidence.
- A to-do list is extracted from a paragraph with no manual tagging.
- A team wants a verifiable context layer with tests that run without an API key.
Not a fit when
- You need a ready planner with a calendar and reminders.
- Notes cannot be stored locally in SQLite.
- Memory must sync across devices automatically.
Example request
Remember that I prefer simple tools and add a commitment to ship the onboarding page next week.MCP
- Transport
- stdio
- Authentication
- not required
Limitations
The last commit dates from September 2026. Without an API key a deterministic demo provider runs, and its text understanding is simpler than a language model. Memory stays in a local file, so backups and file protection are your responsibility. The server keeps memory without a cloud, but any fact you ask it to remember is passed to the agent, so avoid confidential facts. Packages are split into core, MCP and client, and the full API server starts through a separate monorepo command.
How to disable. Remove the entry with `claude mcp remove aldus-palace`. Delete the SQLite memory file separately if the data is no longer needed.
FAQ
Is an API key required?
Not for the demo: `pnpm demo` and the tests run offline on a deterministic provider.
Where is data stored?
In one SQLite file that you own. Any record can be deleted.
Which integration surfaces exist?
MCP, an HTTP API, a library and a typed client all work with the same file.
How can the project be assessed in five minutes?
Open `EVALUATION.md`, which maps every claim to the command that proves it.
Which licence applies?
Apache-2.0.
Discussion
No questions or reviews yet.
Related
MCP serversLocal memory for an agent: hooks record facts from conversations and the MCP server brings them back in later sessions from one SQLite file
LongMemory keeps project memory in local SQLite and serves it to an agent over MCP: decisions, task state, skills and a code graph
Local memory for agents: stores conversations and project files verbatim and retrieves them with semantic search over MCP
Local memory for AI agents: Pensyve keeps facts, conversations and action outcomes in SQLite and returns them in later sessions.
An MCP server that preserves Claude Code context between sessions: decisions, checkpoints and a file cache kept in local SQLite.
A local-first MCP memory server for agents: session history, codebase index and a concept graph in one SQLite database
README in short
The README maps each claim to a verification command, and `EVALUATION.md` lets a reader assess the project in five minutes. Tests and acceptance fixtures run offline.