Self-Hosted Memory Service for AI Agents

Productivity

A memory service for AI agents that runs on your own machine: start it with Docker Compose and get HTTP and MCP interfaces plus a web console, with the vault kept on a local Docker volume. Derived from a frozen snapshot of an MIT-licensed open-source project; ships a Dockerfile, compose file, config template, key-generation script and English quickstart.

Subscribe & download See how to start

SubscriptionUS$49 per month

Last run2026-11-14

In practice

Readiness4 / 5

What you need

Few changes needed4 / 5

Who it is for

Individuals and small teams who use several AI agent tools and want their accumulated notes on their own machine, with a resident local memory service that agents can read and write over HTTP or MCP.

How to start

$ sh scripts/quickstart/init-env.sh
$ docker compose up -d --build

Browse the files

Files in the product repository at the pinned commit. A few house files are left out of the zip.

File overview569 paths

▸assets3

demo.gif
hero.png
loop.svg

▸docs24

README.md
adr/
agents.md
architecture.md
browser-extension.md
cli.md
cli.zh-TW.md
configuration.md
contributing.md
deployment.md
…

▸examples28

demo-vault/

▸extensions16

brain/

▸obsidian_wiki38

__init__.py
__main__.py
ast_extractor.py
batch.py
cache.py
cli.py
client.py
clock.py
code_understanding.py
code_understanding_builtin.py
…

▸scripts8

com.obsidian-wiki.daily-update.plist
daily-update.sh
extract-jsonl.py
manifest.py
obsidian-wiki-daily-update.service
obsidian-wiki-daily-update.timer
quickstart/
wiki-notify.sh

▸skills72

claude-history-ingest/
code-understand/
codex-history-ingest/
copilot-history-ingest/
cross-linker/
daily-update/
graph-colorize/
hermes-history-ingest/
impl-validator/
llm-wiki/
…

▸tests102

conftest.py
fixtures/
test_ast_extractor.py
test_batch.py
test_cache.py
test_cache_manifest_shapes.py
test_claude_plugin.py
test_cli_install.py
test_client.py
test_clock.py
…

▸tools1

check_readme_sync.py

▫AGENTS.md

▫ATTRIBUTION.md

▫CLAUDE.md

▫CONTEXT.md

▫Dockerfile

▫GEMINI.md

▫LICENSE

▫PRODUCT.md

▫QUICKSTART.md

▫README.md

▫README_TW.md

▫SECURITY.md

▫SECURITY.upstream.md

▫SETUP.md

▫delivery-manifest.txt

▫docker-compose.yml

▫pyproject.toml

▫setup.sh

▫uv.lock

▫Config and hidden files 258

Latest passing run

Clone → Checkout → Toolchain → Prepare env and build → Start and fetch home

Pre-purchase Disclosure

What you are buying

A memory service for AI agents brought up with Docker Compose on your own computer: it offers HTTP and MCP interfaces and a web console, and agents read and write a vault kept on your machine using a key. The key is generated by the included script; no third-party service keys are needed.

Derived from an open-source project under the MIT license, and not the original project itself.

Delivery is a frozen snapshot: one zip, taken from the product repository at revision f063fdd.

Known limitations

The service is bound to 127.0.0.1 on the local machine only (port 8080) and is not open to other machines
The vault is stored in a Docker volume, and docker compose down -v deletes it as well
The first start builds the image, and the build needs a network connection to download packages
The dependencies were not scanned for licenses one by one
Pinned to the delivered snapshot; it does not follow upstream.

The liability cap, the scope of what is promised and the statutory floor are governed by the Terms of Service.

Vetting record

Everything below was checked against the pinned commit 7b07334. Every verdict carries its full reasoning, an evidence link and the checker; failures are published, not hidden — do not take our word for it, verify it yourself.

3 PASS2 WARN0 FAIL

Source ReviewWARN

Only part of the delivery was read: docker-compose.yml, .env.example, QUICKSTART.md, Dockerfile and the key-generation script. The Python source of the server was not read file by file.

Script SafetyPASS

Read the Dockerfile and scripts/quickstart/init-env.sh. The script only copies .env.example to .env and fills in a random key; the Dockerfile installs the package with pip and runs as a non-root user. No download-then-pipe-to-shell shape was seen. Other scripts in the repo were not read and are not on the buyer path in the quickstart.

RecencyPASS

This product is a frozen snapshot and does not follow upstream: the upstream source is pinned to the version recorded in the birth-certificate file (see the evidence link). Recent upstream activity was not checked.

LicensePASS

SPDX:MIT

The LICENSE file at the pinned upstream version begins with MIT License and a copyright line, and the birth-certificate file records MIT. The full text was not compared word for word with the standard MIT text. The verdict does not rely on the GitHub license field.

Dependency LicensesWARN

Not verified: the dependency tree (fastapi, uvicorn, mcp, system packages) was not scanned for licenses, and those packages are fetched by the buyer at build time. This entry records that gap, not a scan result.