AI Agents
langmem
LangMem: long-term memory primitives for LangGraph agents that learn and adapt from conversations.
★ 1.7k starsMIT
View guide →Long-term memory runtime for AI agents — plain Markdown as truth, local ranked retrieval, no API key
agent-memory is an open-source ai agents project by tigerless-labs (tigerless-labs/agent-memory on GitHub). It has 2,646 GitHub stars and 142 forks, is released under the MIT license, and was last updated on GitHub on 7 October 2026.
Decide which agents should share long-term memory and where the store should live (default ~/agent-memory-store).
Clone the repo and build the mem, mem-mcp and mem-hook binaries with uv sync --all-packages.
Run mem init, then mem setup --host claude-code / codex / muse-code to merge hooks, skills and MCP entries safely into each host.
Record a decision with mem record, recall it with mem recall, then delete .index and mem rebuild to prove nothing was lost.
One memory = one Markdown file with frontmatter (abstract, type, validity interval, links, provenance) — greppable, git-able and portable.
Recall returns an L0 list of paths/abstracts/scores (via BM25 FTS5, optionally fused with vectors), so the agent opens only the depth it needs.
Distillation triggers at conversation boundaries (SessionStart/Stop/SessionEnd hooks) without the agent having to remember to write.
An independent consolidation pass on its own clock can add and update; deletions only ever arrive as proposals you confirm, and supersede keeps history (recall --as-of).
Claude Code, Codex CLI and Muse Code share one store through the same pipeline; the CLI is the universal fallback for anything that can run a shell command.
There is no LLM client inside the library — judgement is borrowed from the host agent's own CLI, keeping writes visible in your transcript.
Clone agent-memory and build its binaries with uv (Python 3.12+, no PyPI release yet).
git clone https://github.com/tigerless-labs/agent-memory.git
cd agent-memoryCreate the default store at ~/agent-memory-store (or set AGENT_MEMORY_STORE).
mem initProbe and configure a host, then record and recall a first memory.
mem setup --host claude-code
mem record --type decision --field project=agent-memory --abstract "Markdown files are the single source of truth" --body "Indexes are rebuildable caches."tigerless-labs maintains it on GitHub.
Yes, MIT-licensed, and it runs entirely locally.
No API key at all — the README stresses none of it needs one; distillation borrows the host agent's existing login/model.
Those are service/graph or profile systems. agent-memory is a local runtime whose store is plain Markdown files with a rebuildable index, shared across agent hosts.
Clone, uv sync --all-packages, mem init, mem setup --host <your agent>, then mem record / mem recall.
LangMem: long-term memory primitives for LangGraph agents that learn and adapt from conversations.
Multi-agent platform where every agent gets its own computer and long-term memory.
LongMemory: local-first memory for AI agents with temporal truth, provenance, and hybrid recall
Self-managing, unbounded context and long-term memory for coding agents: one session, for life.
Open-source AI memory that gives agents persistent long-term recall.
One portable, local-first memory layer for every AI agent - Markdown-native, user-owned and self-evolving.
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if agent-memory is not the right fit.
A personal AI that really does things — on any OS, with your data staying yours.
An agentic skills framework & software development methodology that works.
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
The agent that grows with you — a Python framework for adaptive AI agents.
DeepSeek's plugin-based agent harness where everything is a plugin.
Free and open — grab the source, star the repo, and start creating.