Home

Recollectium Wiki

Welcome to the Recollectium manual.

Recollectium is a local-first memory engine for AI agents. It gives agents a durable, inspectable place to store and search facts, preferences, decisions, task context, and project knowledge without locking that memory inside one client.

What sets Recollectium apart

Recollectium is not just a helper database for one agent. It is a portable memory layer designed around ownership, semantic recall, and agent interoperability.

Local-first by default: memory lives on your machine unless you choose otherwise.
User memory matters: Recollectium treats user preferences, communication style, durable goals, and personal context as first-class memory, not an afterthought behind project-only notes.
Semantic search by meaning: agents can ask for "preferences about wording" and find the right preference memories without needing an exact type filter every time.
Portable across agents and models: any tool that can use CLI, Python, HTTP, or MCP can use the same Core.
Simple to self-host: installation is one command, runtime is local, storage is SQLite, and services are managed by Recollectium.
Open source and community driven: Recollectium is built for people who want their memory stack to be inspectable, extensible, and owned by the community.

Start here

Quick Start: install, initialize, add memories, search, and start services.
Installation: one-command install, package-manager installs, development install, upgrade, and uninstall.
Concepts: user memory, workspace memory, semantic recall, aliases, and embeddings.
CLI Reference: all commands and flags.
API Reference: local HTTP API calls, parameters, responses, and errors.
WebUI: local browser control plane for managing Core from a browser.
MCP Server: MCP modes and tool contracts.
Troubleshooting: common install, PATH, config, service, API, and MCP fixes.