CPersona Documentation

CPersona is an MCP server that gives Claude — or any MCP-capable agent — persistent memory across sessions. Memories live in a single local SQLite file and are retrieved with a 3-layer hybrid search (vector + FTS5 + keyword, fused by rank or relative score). The server has zero LLM dependency: it never calls a generative model. Two caveats on what that buys you — embeddings can still cost money (EMBEDDING_MODE=api bills per request against an endpoint that defaults to OpenAI's; http mode against a local server does not), and recall is deterministic given a calibrated gate, but the gate itself is measured by sampling the corpus at random, so two installs on identical data can settle on different operating points.

Applies to: CPersona 2.5.x. This site is the canonical documentation — when the README or the bundled skill disagrees with a page here, this site wins, and the discrepancy is a bug worth reporting.

Where to go

You want to…	Read
Install and set up	Getting Started
Know what behaviors you can rely on	Behavior Contracts
See what each of the 30 tools does	Tools
Understand how retrieval and storage work	Architecture
Run it well: backup, tuning, degradation, corpus indexing	Operations Runbook
Look up an environment variable	Configuration
Quick answers to common operator questions	FAQ
Understand a subsystem's design	Design documents (sidebar)
Release tiers and support windows	Release lifecycle + SUPPORT.md

The three memory types

• Declarative — individual facts, decisions, rules (store / recall).
• Episodic — session summaries (archive_episode), which also drive the episode boundary penalty.
• Profile — accumulated user/project attributes (update_profile), with a scoring caveat worth knowing.

For AI agents reading this site

A machine-readable index of these pages is published at llms.txt. The bundled cpersona-memory skill teaches an agent the day-to-day store / recall / archive workflow and links back here for the canonical detail.