Pensyve MCP
Local memory for AI agents: Pensyve keeps facts, conversations and action outcomes in SQLite and returns them in later sessions.
- 100 ★ on the repository
- Last commit
Extended access
Why this level
- Stores the content of conversations and project facts in a local database, so restrict access to it.
- An HTTP gateway exposed beyond localhost needs your own key and HTTPS.
Install
- Works in
Manual install
git clone https://github.com/major7apps/pensyve.git
cd pensyve
cargo install --path pensyve-mcp --lockedBuilds the MCP server from source; needs Rust 1.94 or newer.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Stores the content of conversations and project facts in a local database, so restrict access to it.
- An HTTP gateway exposed beyond localhost needs your own key and HTTPS.
This is third-party code. Review the repository files before installing.
What it does
Pensyve is a persistent memory engine for agents. The pensyve-mcp server gives Claude Code, Codex and Cursor tools such as `pensyve_remember`, `pensyve_recall`, `pensyve_observe` and `pensyve_inspect`. The Rust engine keeps facts, conversations, action outcomes and observations outside the language model, writes them to local SQLite and runs embedding models on your machine, so search works by meaning. Local use needs no account and no API key. The same engine is reachable from Python, the command line and a REST gateway.
Who it is for. Developers who want Claude Code or Cursor to remember project agreements between sessions without cloud storage.
Good fit when
- An agent forgets the library choice and code rules agreed last week.
- Several projects keep separate memory through different PENSYVE_NAMESPACE values.
- A Python agent saves a fact about a user and retrieves it in the next process.
- A team runs a shared HTTP gateway and connects TypeScript or Go clients to it.
Not a fit when
- You want memory in a hosted service: Pensyve Cloud closed on October 1, 2026.
- You expect new features: the project is in maintenance mode with security fixes and dependency updates only.
- You plan to fine-tune a model: saved records do not change its weights.
Example request
Remember that this project uses pnpm and strict TypeScript, and remind me of it in the next session before I edit the configuration.MCP
- Transport
- stdio, http
- Authentication
- not required
Environment variables
PENSYVE_NAMESPACE- Memory namespace, one per project.
Limitations
Pensyve Cloud closed on October 1, 2026, and the project lives on as open source in maintenance mode: security fixes and dependency updates, no new features. The MCP server is built from source with `cargo install`, so it needs Rust 1.94 or newer and the Cargo binary directory on the client PATH. Embedding models may download on first start, so prepare the files in advance for offline use. TypeScript and Go clients connect only to an HTTP gateway you run yourself. Your integration decides what to save and when: a recorded conversation does not turn into a successful procedure on its own.
How to disable. Remove the pensyve entry from the client MCP configuration and the pensyve-mcp binary. Data sits in local SQLite storage, so delete it separately.
FAQ
Does Pensyve need an API key?
No, the local MCP server works without a key or account. Only the HTTP gateway you run yourself needs a key with the psy_ prefix.
How do I connect Pensyve to Claude Code?
Install pensyve-mcp with `cargo install` and add a stdio entry with the command `pensyve-mcp` and the `--stdio` argument. The repository also ships a Claude Code plugin described in integrations/claude-code.
Does the Cloud shutdown break the local version?
No. The engine, SDKs and MCP server stay open and need no cloud. A saved Cloud export can be loaded into your own gateway following the self-hosting guide.
Does Pensyve train the model on saved data?
No. Memory lives outside the model, and the application retrieves relevant records and adds them to a later prompt.
Which tools does the MCP server have?
They include `pensyve_remember`, `pensyve_recall`, `pensyve_observe` and `pensyve_inspect`. The first two save and search memory, and the rest are described in docs/GETTING_STARTED.md.
Discussion
No questions or reviews yet.
Related
MCP serversA local-first MCP memory server for agents: session history, codebase index and a concept graph in one SQLite database
MenteDB keeps agent memory in a Rust database: an MCP server for Claude Code and Cursor in cloud or self-hosted mode.
A long-term memory MCP server for agents: one Rust binary that runs offline and never calls a language model
Local memory for agents with a review queue: SQLite, an audit log and a token ledger, one vault for Claude Code, Codex and Cursor
Local memory for agents in a single Rust binary: stores, searches by meaning and keyword, works with no network and no keys
MemMesh, MCP-сервер памяти в одном бинарнике на Rust, хранит заметки агента на машине и ищет по смыслу без внешней базы.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Pensyve (this card) | no | not required | 100 | |
| Mem0 | no | 68,887 | a memory layer with plugins and SDKs; Pensyve stores everything locally in SQLite and ships a Rust MCP server | |
| Basic Memory | no | not required | 4,557 | keeps memory in Markdown files; Pensyve writes facts, conversations and action outcomes to SQLite |
README in short
Four kinds of records: facts, conversations, action outcomes and observations. Interfaces: Python SDK, MCP server, the pensyve CLI and a REST gateway on SQLite or PostgreSQL. After the cloud version closed, the project moved to maintenance mode.