OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Agent Memory & RAG

Agent Memory

Agent Memory is a Python-based local memory runtime for AI agents. It stores each memory as a Markdown file, maintains a rebuildable SQLite index for ranked retrieval, and includes a separate sleep-time layer for consolidating accumulated memories. Its command-line setup supports Claude Code, Codex CLI and Muse Code.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Agent Memory — OpenViST project profile in Agent Memory & RAG.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Agent Memory when an agent needs cross-session recall and you want the underlying knowledge kept in ordinary Markdown files, with local ranked search and documented host setup.

Who it is for

  • Developers configuring persistent memory for coding agents.
  • Users who want memory files to remain inspectable and portable rather than locked in an opaque store.
  • People running supported agent hosts that can use the command-line runtime.

Documented uses

  • Keep useful knowledge available after an agent session ends.
  • Find relevant memories locally and open only the needed file or level of detail.
  • Share a memory store across Claude Code and Codex CLI.
  • Manage accumulated memories through a separate consolidation process.
Before you choose

Requirements & limits

The repository is public and lists an MIT license. The runtime is described as requiring no API key; use requires a local command-line setup and Python 3.12 or later.

  • Markdown files are the source of truth; the SQLite index is a rebuildable cache.
  • Retrieval returns paths and ranked results rather than automatically pasting full memory text into the agent context.
  • The README documents lifecycle-hook setup for Claude Code, Codex CLI and Muse Code.
  • A full session trace is copied before distillation, and deletion during sleep-time management is presented as a proposal for confirmation.
  • Core retrieval is local; the README describes an optional vector plugin fused with BM25 using RRF.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up Agent Memory for an agent host

Initialize the store and configure a supported host integration.

Before you start

  • Python 3.12 or later.
  • A command-line environment with Agent Memory installed.
  • A supported host: Claude Code, Codex CLI or Muse Code.

Steps

  1. Run mem init to initialize the store; it defaults to ~/agent-memory-store.

  2. If using a different store location, set AGENT_MEMORY_STORE to its absolute path in the environment where the agents run.

  3. Run mem setup --host claude-code, mem setup --host codex, or mem setup --host muse-code as appropriate; Muse Code setup may include a provider option.

  4. Run the corresponding mem doctor --host ... command to inspect host, configuration, store, hook and reasoner status.

Expected outcome

The host setup process installs lifecycle hooks and the skill, configures host settings and performs preflight; a successful setup reports `status: READY`.

Guide limitations

  • Setup and doctor perform a minimal live reasoner request by default; --no-live performs offline inspection but reports FAILED because readiness was not proven.
  • Codex user hooks remain subject to its trust prompt.
Documentation 1 ↗
Source-based practical guide

Record a memory and rebuild its index

Create a Markdown-backed memory, retrieve it, and rebuild the local index.

Before you start

  • Agent Memory installed in a command-line environment.
  • An initialized store.

Steps

  1. Run mem record with a type, field, abstract and body to add a memory.

  2. Run mem --json recall with a query to retrieve matching results.

  3. Remove the store's .index directory and run mem rebuild to recreate the index from the Markdown files.

Expected outcome

The memory is stored as a file and can be found through local retrieval; the index can be regenerated from the file-based store.

Guide limitations

  • The repository’s example demonstrates one memory and one query; it does not establish retrieval quality for other stores or workloads.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
9 October 2026
Original website title
tigerless-labs/agent-memory · Official homepage title · retained in its source language
How we review sourcesSuggest a correction