# FluctlightDB

> FluctlightDB is an embedded memory database for AI agents. experience() writes, activate() recalls from a cue, checkpoint() keeps it. Open source, beta.

An embedded memory database for AI agents. A brain is a directory, the engine
runs in your process, and three calls make up the contract: experience() writes,
activate() recalls from a cue, checkpoint() makes it survive a restart.

Install: pip install "fluctlightdb[native]"

## Read this first

- [Full documentation, plain text](/llms-full.txt): every signature, every
  return shape, and the six mistakes agents make against this API. Start here.

## Pages

- [If you are an agent](/docs#agents)
- [Mistakes against this API](/docs#mistakes)
- [Install](/docs#install)
- [Your first brain](/docs#first-brain)
- [What a brain is on disk](/docs#on-disk)
- [experience() — writing](/docs#experience)
- [activate() — recall](/docs#activate)
- [Recall is lexical offline](/docs#lexical)
- [Provenance and trust](/docs#provenance)
- [checkpoint() and durability](/docs#checkpoint)
- [Which entry point](/docs#entry-points)
- [Stable and experimental](/docs#surface)
- [Reference](/docs#reference)

## On this site

- [Install](/install): the pip line, the canonical sample and the checkpoint note.
- [Quickstart](/quickstart): the install line and the canonical sample.
- [Paper](/paper): the preprint, a citation and BibTeX.
- [Benchmarks](/benchmarks): the figures this site stands on, with conditions.
- [Glossary](/glossary): engram, provenance, cue recall, consolidation.
- [Changelog](/changelog): GitHub releases. Not a version to pin.
- [Questions, answered](/faq)
- [Compared with Mem0, Zep, Letta, LangMem and vector databases](/compare)
- [What is wrong with it](/catalogue#condition)

## Integrations, use cases, notes

- [One brain, several ways in](/integrations)
- [FluctlightMemory, and a chat history that returns the session](/integrations/langchain)
- [FluctlightLlamaMemory constructs, on main](/integrations/llamaindex)
- [Two handlers, and no SDK in the process](/integrations/openai-agents)
- [A memory server for Cursor, Claude and Codex](/integrations/mcp)
- [The image the repository publishes](/integrations/docker)
- [HTTP from Node, against a server this page did not start](/integrations/node)
- [Four things the engine was asked to do](/use-cases)
- [An agent that remembers a project decision](/use-cases/coding-agent)
- [A ledger line outranks a line from chat](/use-cases/provenance)
- [A long run that can be killed](/use-cases/crash-safe)
- [An agent with no server and no embedder](/use-cases/offline)
- [Mem0 is a memory layer. This is the database under one.](/compare/mem0)
- [Zep tracks facts over time. This stores episodes.](/compare/zep)
- [Letta is where the agent lives. This is storage it can call.](/compare/letta)
- [LangMem is LangChain’s memory. This is a database that also plugs in.](/compare/langmem)
- [Nearest is one signal. It is not the question.](/compare/vector-databases)
- [SQLite plus embeddings is still SQL](/compare/sqlite)
- [Three notes, from the repository](/notes)
- [A memory without a source is hearsay](/notes/provenance)
- [The headline that counted turns the engine never retrieved](/notes/locomo)
- [Kill the writer. Open the brain. See if it verifies.](/notes/sigkill)
- [Description, citation, assets, contact](/press)
- [Atom feed of the notes](/feed.xml)

## For machines

These are generated from the same sources as the HTML. They are not in the sitemap.

- [Facts](/facts.json): name, install line, verbs, stable surface, limits, benchmarks with conditions, and the retracted figure named as retracted.
- [Home, Markdown](/index.md)
- [Documentation, Markdown](/docs.md)
- [Quickstart, Markdown](/quickstart.md)
- [Install, Markdown](/install.md)
- [Questions, Markdown](/faq.md)
- [Comparison, Markdown](/compare.md)

## Elsewhere

- [Source](https://github.com/voxmastery/FluctlightDB)
- [Package on PyPI](https://pypi.org/project/fluctlightdb/)
- [Paper (preprint)](https://arxiv.org/abs/2608.12365)
- [Paper on Hugging Face](https://huggingface.co/papers/2608.12365)
- [Benchmarks, in the repository](https://github.com/voxmastery/FluctlightDB/blob/main/docs/BENCHMARKS.md)

## Conditions

Beta software, one maintainer, no third-party security audit. Benchmark figures
are maintainer-reported and have not been independently reproduced. The retracted
99% LoCoMo figure is not a current claim; see /catalogue#measurements.