AI Agents
memvid
A serverless, single-file memory layer for AI agents - replace complex RAG pipelines
★ 16.6k starsApache-2.0
View guide →// agents · FlowElement-xinliuyuansu/m_flow
m_flow
m_flow is an open-source ai agents project by FlowElement-xinliuyuansu (FlowElement-xinliuyuansu/m_flow on GitHub). It has 4,514 GitHub stars and 258 forks, is released under the Apache-2.0 license, and was last updated on GitHub on 1 September 2026.
M-flow is a bio-inspired cognitive memory engine for AI agents — a graph alternative to flat RAG
★ 4.5k stars⑂ 258 forksApache-2.0updated 2026-09
▦ visual — this project's README ships no embeddable screenshot,
so here's a placeholder. The guide below is complete regardless.
so here's a placeholder. The guide below is complete regardless.
// project facts
Project facts
- Repository
- FlowElement-xinliuyuansu/m_flow
- GitHub stars
- 4,514
- Forks
- 258
- License
- Apache-2.0
- Last updated on GitHub
- 1 September 2026
- Category
- AI Agents
- Listing
- Full install guide
- Official site
- Official site
// how it works
How it works
- Decide your data
Gather the conversations, documents, or files you want the memory engine to learn from.
- Install M-flow
Install from PyPI with pip or uv.
- Configure an LLM
Point M-flow at an LLM (for example an OpenAI-style API) used for ingestion and query.
- Ingest and query
Add your data, then ask a question and compare episodic versus triplet-completion retrieval.
// features
Why people love it
Four-layer cone graph
Episodes decompose into facets, facet points, and entities, so retrieval can start at the right level of abstraction.
Evidence-path scoring
Answers are ranked by the strongest chain of weighted evidence, not just embedding similarity.
Coreference-aware ingestion
Pronouns and references are resolved at write time, so related facts join the same chains.
Five retrieval modes
Episodic, Procedural, Triplet Completion, Lexical, and Cypher modes for different question types.
Procedural memory
Remembers how tasks were done, not only facts that were stated.
Broad ingestion and MCP
50+ file formats supported, plus MCP integration for agent clients.
Advertisement
// run it
Run it yourself
- Clone the repository
Clone the project and enter its directory for a source install.
git clone https://github.com/FlowElement-xinliuyuansu/m_flow.git cd m_flow - Install from PyPI
Install the published package (or `pip install -e .` from the clone).
pip install mflow-ai - Ingest and query
Use the Python API documented in the README to add data and run your first query.
# see README for the exact ingest/query API calls
// use cases
What you can build
- Give a personal agent long-term memory that connects related facts across months of chats
- Answer 'who decided what, and why' questions over meeting history
- Build a research assistant that traces claims to their strongest evidence
- Add graph-based memory to an MCP-compatible agent stack
Advertisement
// faq
Questions, answered
Who maintains M-flow?
FlowElement (flowelement.ai) develops M-flow as an open-source project.
Is M-flow free?
Yes, Apache-2.0 licensed. You provide the LLM used for ingestion and querying.
Does it need a paid API key?
It needs an LLM for ingestion/query — typically your own API key for an OpenAI-style provider, as configured per the README.
How does it compare to GraphRAG or a vector store?
Vector stores retrieve similar chunks; GraphRAG builds entity graphs. M-flow's cone graph adds graded abstraction layers and scores answers by evidence-path strength with coreference resolved at ingestion.
How do I get started?
Run `pip install mflow-ai`, configure your LLM, ingest a small dataset, and try the Episodic and Triplet Completion modes from the README examples.
// keep exploring
Similar tools
AI Agents
Memary
The open-source memory layer for autonomous agents — knowledge-graph memory that emulates human memory
★ 2.7k starsMIT
View guide →AI Agents
MemOS
Self-evolving memory OS for LLMs and agents: persistent graph memory, hybrid retrieval, skill reuse.
★ 11.8k starsApache-2.0
View guide →Agents
danghuangshang
Multi-agent court inspired by ancient China: specialized agents debate and decide.
★ 2.7k starsMIT
View guide →AI Agents
ReMe
A memory management kit for agents — remember me, refine me.
★ 3.6k starsApache-2.0
View guide →Agents
agentic-context-engine
Make your agents learn from experience with a dedicated context engine.
★ 2.6k starsApache-2.0
View guide →// alternatives
Alternatives in AI Agents
The most-starred other AI Agents projects in the LibreAI catalogue — the realistic shortlist if m_flow is not the right fit.
- openclaw★ 391.6k
A personal AI that really does things — on any OS, with your data staying yours.
- superpowers★ 294.9k
An agentic skills framework & software development methodology that works.
- ECC★ 275.2k
Agent harness optimization system with skills, memory and security for Claude Code, Codex and more
- hermes-agent★ 252.0k
The agent that grows with you — a Python framework for adaptive AI agents.
- deepseek-harness★ 245.5k
DeepSeek's plugin-based agent harness where everything is a plugin.
// more in agents
More in AI Agents
Nearby AI Agents projects in the catalogue, closest in stars to m_flow — keep exploring the shelf either side of it.
- LongMemory★ 4.5k
LongMemory: local-first memory for AI agents with temporal truth, provenance, and hybrid recall
- logfire★ 4.5k
Pydantic Logfire: Python and AI observability on OpenTelemetry — SQL-queryable traces, metrics, and logs
- council-of-high-intelligence★ 4.6k
Convene 18 legendary thinkers as an AI council that debates your hardest questions — inside Claude Code
- embabel-agent★ 4.5k
Agent framework for the JVM. Pronounced Em-BAY-bel /ɛmˈbeɪbəl/
- PyRIT★ 4.6k
Microsoft's open framework for red-teaming generative AI systems.
- memory-lancedb-pro★ 4.5k
Memory-LanceDB-Pro gives OpenClaw agents hybrid search, smart extraction, and intelligent forgetting
Get m_flow
Free and open — grab the source, star the repo, and start creating.