Aletheia Mem

9 stars
Apache-2.0
Active this week
Last commit 1 day ago

Last verified Oct 31, 2026

Manage shared memory for AI coding tools in one SQLite store over MCP, REST or CLI, with rules synced to each tool's instruction file. Apache-2.0, no account.

Visit website GitHub9 Deploy on Hostinger

Read the full breakdownWho it is for, how it works, trade-offs, the quick start and the FAQ

who it's for

Who Aletheia Mem is for#

Developers using two or more AI coding tools who want shared context

If you work with Claude Code for some tasks and Codex or Cursor for others, Aletheia keeps one memory store that all of them read. You do not need to re-establish project context when switching tools. Rules written once apply everywhere once synced into each tool's instruction file.

Skip if:

If you use only one AI coding tool and do not switch between them, a shared store adds no value. Per-tool memory and instruction files are sufficient.

Privacy-conscious developers avoiding cloud memory services

All data stays in a local SQLite file. The server listens on loopback (127.0.0.1) by default. No account or telemetry. If you want persistent context across sessions without routing any of it through external APIs or cloud services, this is the primary use case the tool is designed for.

Skip if:

If you are comfortable with cloud-hosted memory and do not have data-residency concerns, managed services may be simpler to set up and maintain without running a local server.

Claude Code users who want cross-session memory and rules sync

Claude Code's integration is the most thoroughly tested: hooks ran in real sessions over two days of use on one machine. aletheia install claude-code --write sets up the MCP entry, hooks and rules block in one command, with previews and backups before any write.

Skip if:

If claude-mem already covers your Claude Code memory needs and you do not need to share memory with other tools, migrating has limited upside at this early stage of the project.

the problem

The problem it solves#

Every AI coding tool forgets between sessions. Claude Code, Codex, Cursor, and Gemini CLI each keep their own state, and none of it carries across tools or sessions. A rule set in Claude Code does not exist in Codex. A fact surfaced in one session is invisible to every other tool the next time you open it.

Developers working across more than one AI coding tool repeat the same context setup every time they switch. There is no standard way for a tool that learned something yesterday to share it with a different tool today, and no central place to write rules once and have them propagate to every instruction file in your stack.

how Aletheia Mem solves it

How it solves it#

Single SQLite store over MCP, REST and CLI

All AI coding tools read from and write to one file at ~/.aletheia/hub.db. Any tool that speaks MCP stdio or HTTP, calls the REST API at /v1, or uses the CLI can access the same memory. The store requires no external database and no cloud account.

Rules sync to 9 instruction file targets

Write a rule once and sync it into CLAUDE.md, AGENTS.md, GEMINI.md, Cursor's .mdc rules, Windsurf's rules, Copilot instructions, and Aider files. Each target gets a managed block, so your own text is never overwritten. The sync is previewed before any write, and every changed file gets a timestamped backup.

Automatic hook capture in Claude Code, Codex and Command Code

Hooks installed via aletheia install capture SessionStart, PostToolUse, Stop and other events. Secrets and <private> blocks are removed before anything is stored. If the database is temporarily unavailable, hooks spool to disk and exit 0 so they never block the host tool. Hooks have been tested in real sessions of Claude Code (two days of use), Command Code (one short session) and Codex CLI (one interactive session), all on one machine and one user.

Import from claude-mem

aletheia import claude-mem reads an existing claude-mem database in read-only mode and brings your memories across. The import is idempotent: running it twice does not duplicate entries. Aletheia runs on its own port and never writes to claude-mem's database, so both can run simultaneously.

Private by default with secret redaction

Hooks remove secrets and <private> blocks from every entry before storage. Optional LLM features are off until you configure a provider. The server listens on loopback only; no data leaves your machine unless you explicitly turn on a model provider or share an export.

strengths · trade-offs

Strengths and trade-offs#

Strengths

Works with any tool speaking MCP or reading instruction filesAletheia does not depend on any one AI coding tool. Any tool that speaks MCP stdio or HTTP, or that reads an instruction file, can connect to it. The README capability table shows configuration support for Claude Code, Codex, Cursor, Gemini CLI, Windsurf, OpenCode and Command Code, with verified MCP connections for Claude Code, Command Code, Codex and OpenCode as of 2026-10-27.
Apache-2.0 with no account, no telemetry and no hosted tierThe license is Apache-2.0, which permits self-hosted use, commercial internal use, and modification without restriction. There is no managed tier, no usage fee, and no account to create. All data stays on your machine by default; data leaves only through a model provider you configure or an export you explicitly share.
Reproducible retrieval benchmark with measured numbersThe project publishes retrieval benchmarks on the LoCoMo dataset (1,536 questions) with all result files in bench/results/ and instructions to rerun them. Hybrid search with dataset facts reaches Recall@5 of 64.8% at 5.53 ms/query on the full 1,536-question set. The measured baseline for claude-mem 15.2.0 as shipped is Recall@5 36.0% at median 1,699 ms/query. These are retrieval numbers, not answer accuracy.

Trade-offs

-New alpha project with limited real-session verificationAletheia Mem was created on 2026-10-26 and carries alpha status in the README. Hook capture has been tested in real sessions only for Claude Code, Command Code and Codex CLI, all on one machine and one user. Cursor, Gemini CLI, Windsurf and OpenCode hooks have not run in a real session. Answer accuracy is not measured. Features may change without notice.
-Requires Bun 1.1 or newer rather than Node.jsThe server and CLI run on Bun, not Node.js. Developers who do not have Bun installed need to add it before the tool will work. The npm install step works with any package manager, but the installed binaries invoke Bun at runtime; Node alone will not run them.
-Hook testing limited to one machine and one developerAll real-session hook testing was performed by the author on their own machine. The project distinguishes clearly between 'config written from docs' (Cursor, Gemini CLI, Windsurf) and 'hooks ran in real sessions' (Claude Code, Command Code, Codex). The tool has not been vetted in multi-user or multi-machine setups.

versus alternatives

Aletheia Mem vs alternatives#

Aletheia Mem vs Mem0

Both tools store memory for AI agents, but they serve different deployment models. Mem0 offers a managed cloud API with paid team features alongside an open source self-hosted tier; Aletheia Mem is local-only with no managed tier.

Feature	Aletheia Mem	Mem0
License	Apache-2.0	Apache-2.0 (OSS tier)
Self-hosting	Yes, local only	Yes (OSS tier)
Managed/hosted tier	No	Yes (paid)
MCP	Yes (stdio and HTTP)	Yes
REST API	Yes (/v1)	Yes
CLI	Yes	Partial
Rules sync to tool files	Yes (9 targets)	No
Hook capture	Yes (Claude Code, Codex, Command Code tested)	No
Vector search	Partial (optional ONNX)	Yes
Team sharing	Partial (self-hosted sync, loopback tested only)	Yes (paid)
Published benchmark	Yes, reproducible	Vendor-claimed only

Cells for Mem0 are from Mem0's own public documentation as fetched on 2026-10-26 and recorded in the project's README capability table.

Aletheia is the better fit when you need local-first memory with no data leaving your machine, rules synced automatically into CLAUDE.md and other instruction files, or one shared store across multiple AI coding tools. The self-hosted open source tier of Mem0 does not include rules sync to tool instruction files or hook-based session capture.

Mem0's managed cloud API is worth considering when you need team memory with no infrastructure to run, vector search with graph and entity memory out of the box, or a larger community with more documented production deployments. Aletheia is in alpha, with real-session testing limited to one developer on one machine; teams that need production-grade memory with vendor support should evaluate Mem0's commercial tier first.

install · quick start

Quick start#

bash

Install the package with npm (Bun 1.1 or newer is required as the runtime), then start the server and connect Claude Code.
```bash
npm install -g aletheia-mem
aletheia-server
aletheia install claude-code --write
```

tech stack · detected from GitHub

What it's built on#

Languages: JavaScript
Databases: SQLite
Runtimes: Bun

frequently asked

FAQ#

Does Aletheia Mem replace claude-mem?

No. Aletheia runs on its own port (37790) and never writes to claude-mem's database. The two run side by side. aletheia import claude-mem reads your existing claude-mem memory in read-only mode to bring it across; running the import twice does not duplicate entries. The README states that claude-mem is much larger with more features and a bigger community; Aletheia's differentiation is tool-neutrality and rules sync to multiple instruction files.

Is Aletheia Mem free with no paid tier?

Yes. The license is Apache-2.0, there is no hosted tier, no account, and no usage fee. Optional LLM features call the model you configure locally and are off by default. The only cost is the Bun runtime and a machine to run the server on.

Which AI coding tools have been tested on a real install?

As of 2026-10-27, MCP connections were verified with the real CLI for Claude Code (2.1.312), Command Code (1.88.1), Codex (0.130.0) and OpenCode (1.18.33). Hooks ran in real sessions for Claude Code (two days of use), Command Code (one session) and Codex (one session), all on one machine and one user. Cursor, Gemini CLI and Windsurf are configured from documentation but have not been tested on a real install.

Where does my data live, and can it leave my machine?

All data is stored in ~/.aletheia/hub.db on your machine. The server listens on loopback (127.0.0.1) by default. Data leaves only if you configure a model provider (which sends relevant context to that provider's API) or share an export bundle manually. There is no background sync or telemetry.

Can I share memory across machines or with a team?

Aletheia Sync is a self-hosted, append-only sync server documented in the project. It supports device pairing, optional end-to-end encryption of payloads (which does not hide metadata such as device ID, operation kind or timing), and scoped tokens. The README notes that Sync has been tested on loopback only and not across two real machines. TLS is delegated to a reverse proxy; OAuth is not built in.

also worth a look

Similar open-source tools#

Hermes Agent

Featured

Alternative toGrok Bot

Self-hosted AI agent: persistent memory, multi-channel chat, model choice.

258,472 GitHub stars
Licence MIT

Mlx Serve

Alternative toLM Studio

Local LLM server for Apple Silicon, faster than LM Studio

2,206 GitHub stars
Licence MIT

TensorFold

InAI & Machine Learning

Local LLM inference for Apple Silicon and NVIDIA GPUs, no Python.

1,567 GitHub stars
Licence Apache-2.0

Skillshare

InDeveloper Tools

AI coding setup manager for skills, agents, MCP, and hooks

2,872 GitHub stars
Licence MIT

Agent Desktop

InAI & Machine Learning

Desktop computer use for AI agents via OS accessibility tree

3,850 GitHub stars
Licence Apache-2.0

Claude Cookbooks

InAI & Machine Learning

Official Claude API notebooks and code recipes for Python developers

53,978 GitHub stars
Licence MIT

by category

Open source tools in these categories#

517

Self-host Aletheia Mem on a Hostinger VPS

Keep it running 24/7, even when your laptop is off. A Hostinger VPS gives it a server of its own.

01
Choose a plan
Starts from $6.49/mo
02
Apply
Extra 10% off yearly plans
03
Install Aletheia Mem
On a server that stays on
04
Start building
Focus on the app, not the server

Deploy on Hostinger

Inbox Zero

Alternative toSuperhuman

AI email assistant that drafts replies and clears your inbox

12,855 GitHub stars
Licence AGPL-3.0
#Email

rakazo

Alternative toGrok Bot

AI teammates you own: your keys, your model, your machine.

6,242 GitHub stars
Licence Apache-2.0
#AI Agents

SnitchFeed

Live monitoring on LinkedIn, X, Reddit & more. Agentic social listening for startups.

Marketing & Growth

Advertise your product

The projects we add and every new guide. Free, unsubscribe any time.

Find open source alternatives to paid software

Aletheia Mem vs Mem0

Feature	Aletheia Mem	Mem0
License	Apache-2.0	Apache-2.0 (OSS tier)
Self-hosting	Yes, local only	Yes (OSS tier)
Managed/hosted tier	No	Yes (paid)
MCP	Yes (stdio and HTTP)	Yes
REST API	Yes (/v1)	Yes
CLI	Yes	Partial
Rules sync to tool files	Yes (9 targets)	No
Hook capture	Yes (Claude Code, Codex, Command Code tested)	No
Vector search	Partial (optional ONNX)	Yes
Team sharing	Partial (self-hosted sync, loopback tested only)	Yes (paid)
Published benchmark	Yes, reproducible	Vendor-claimed only

Cells for Mem0 are from Mem0's own public documentation as fetched on 2026-10-26 and recorded in the project's README capability table.

Who Aletheia Mem is for#

Developers using two or more AI coding tools who want shared context

Skip if:

Privacy-conscious developers avoiding cloud memory services

Skip if:

Claude Code users who want cross-session memory and rules sync

Skip if:

The problem it solves#

How it solves it#

Single SQLite store over MCP, REST and CLI

Rules sync to 9 instruction file targets

Automatic hook capture in Claude Code, Codex and Command Code

Import from claude-mem

Private by default with secret redaction

Strengths and trade-offs#

Strengths

Trade-offs

Aletheia Mem vs alternatives#

Aletheia Mem vs Mem0

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

Developer Tools

AI & Machine Learning

LLMOps & AI Tooling

Find open source alternatives to paid software

Who Aletheia Mem is for#

Developers using two or more AI coding tools who want shared context

Skip if:

Privacy-conscious developers avoiding cloud memory services

Skip if:

Claude Code users who want cross-session memory and rules sync

Skip if:

The problem it solves#

How it solves it#

Single SQLite store over MCP, REST and CLI

Rules sync to 9 instruction file targets

Automatic hook capture in Claude Code, Codex and Command Code

Import from claude-mem

Private by default with secret redaction

Strengths and trade-offs#

Strengths

Trade-offs

Aletheia Mem vs alternatives#

Aletheia Mem vs Mem0

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

Developer Tools

AI & Machine Learning

LLMOps & AI Tooling

Find open source alternatives to paid software