docs

Conare Docs

Conare is persistent memory for coding agents. It imports past Claude Code, Codex, and Cursor sessions, exposes them through MCP, and lets agents recall project context instead of starting blank.

Quickstart

Run the installer:

shell
$ npx conare@latest

The wizard signs you in, imports local chat history, detects AI clients, configures MCP, installs the Conare skill, and can enable background sync.

Use Conare in an agent

At the start of a new conversation, call recall once with the current task. Conare returns relevant prior decisions, file paths, commands, preferences, and context from previous work.

During the conversation, call search when the user mentions something unknown or asks for prior work. Do not ask the user to explain a project, person, decision, or command until Conare has searched memory first.

Core pages

NeedPage
Install Conare and connect an MCP clientInstall
Teach an agent how to use memory wellMCP Tools
Run imports, codebase indexing, or sync manuallyCLI
Build a custom clientHTTP API

Concepts

Memories are stored text records. They can be chats, codebase files, imported docs, or user-saved notes.

Containers group memories by source. Common containers are claude-chats, codex-chats, cursor-chats, codebase, and saved.

Recall is for conversation startup. Agents call it once with the current task so Conare can return relevant project history, preferences, and decisions.

Search is for specific lookups mid-conversation. Use it when the user asks about previous work, mentions an unknown project, or needs an old decision recovered.

Defaults

The MCP endpoint is:

https://conare.ai/mcp

API keys start with cmem_ and are sent as bearer tokens:

Authorization: Bearer cmem_...