MnemoSeed

MnemoSeed

One memory for every AI agent you use.

License: Dual AGPL/Commercial CI

Your Claude forgets everything when you open Grok. Your Cursor has no idea what Codex did yesterday. MnemoSeed is a local memory layer that sits underneath all of them. It captures what matters while you work, consolidates it in the background, and injects the right context into whichever agent you open next.

# install once, works everywhere
uv tool install mnemoseed
mnemoseed up        # local daemon, zero Docker needed
mnemoseed install   # wires up Claude Code, Cursor, Codex, Grok Build

How it works

  1. Capture — a deterministic scorer decides what is worth keeping (no LLM call, no API tokens). Logs and noise get stripped, durable facts get stored verbatim.
  2. Dream — when enough has accumulated, a background pass consolidates raw chunks into a structured knowledge graph. Interrupt it any time; nothing blocks your session.
  3. Recall — the next agent (any agent) gets a small, budgeted context package. Hard token cap, conflict pairs returned as-is, and an honest "I don't have anything" when there's nothing.

Everything is local-first. A typical dream run costs about $0.001 against cloud models, and $0 on the offline track.

What you get

  • Cross-model memory — Claude Code, Cursor, Codex, Grok Build, and anything that speaks MCP
  • A graph you can audit — every fact carries provenance (who said it, when, from which session), and history is never overwritten
  • Cost you can predict — dynamic budget per dream (5k–32k tokens, sized by actual backlog) plus a monthly token ledger with a hard cap
  • Privacy by default — local-first by default; encrypted transport and encrypted at-rest storage wherever the daemon runs, and cloud dreams only ever leave through zero-retention (ZDR) model endpoints

Status

Shipped: capture pipeline, dream engine, hybrid retrieval, six MCP tools (memory.recall / remember / audit / timeline / export / forget_this), installer + doctor + uninstall, Claude Code plugin, Cursor / Codex / Gemini adapters. ~1,100 tests green on every push.

Roadmap: management console, hosted cloud daemon (running in a TEE as standard), the anima personality module (spec in docs/design/09).

Design docs live in docs/design/ (English) and docs/zh/ (中文工作稿).

Development

Test-driven, with an adversarial verifier on every task: failing tests first, mutation spot-checks before any merge. Gates: uv run pytest -q, ruff check, ruff format --check, mypy src.

License

Dual-licensed: AGPL-3.0 (free, including commercial use, as long as derivatives stay open) or a commercial license for proprietary/closed-source integration. Contact license@mnemoseed.com.

Key dates

PyPI data
Data sourced directly from PyPI's database.
  • Released:
Newer release available (0.1.1)

1 maintainer

PyPI data
Data sourced directly from PyPI's database.
Avatar for kumhou from gravatar.com kumhou

License

AGPL-3.0-only

Requires

Python >=3.12

Report project as malware

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

mnemoseed-0.1.0.tar.gz (945.9 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

mnemoseed-0.1.0-py3-none-any.whl (759.8 kB view details)

Uploaded Python 3

File details

Details for the file mnemoseed-0.1.0.tar.gz.

File metadata

  • Download URL: mnemoseed-0.1.0.tar.gz
  • Upload date:
  • Size: 945.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mnemoseed-0.1.0.tar.gz
Algorithm Hash digest
SHA256 fb7d4b39c86089d52d6ba769c0c61f88f191dd6b773e7daa2e888ba2ed0ceef7
MD5 dd747280bf59efa7c63468d8a366ee58
BLAKE2b-256 a561c25d57ebca5473ed4cc2d9e7bb230e57def005494e350303e2653f486b46

See more details on using hashes here.

Provenance

The following attestation bundles were made for mnemoseed-0.1.0.tar.gz:

Publisher: release.yml on MnemoSeed/mnemoseed

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file mnemoseed-0.1.0-py3-none-any.whl.

File metadata

  • Download URL: mnemoseed-0.1.0-py3-none-any.whl
  • Upload date:
  • Size: 759.8 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/7.0.0 CPython/3.13.14

File hashes

Hashes for mnemoseed-0.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 6f16944e12809600f44e2ee74bc6cad748dbb9837cabe26bb1a0151c018e4e3e
MD5 0a78db00688363edd105718f627e4fa3
BLAKE2b-256 e70f7c49e81bea04c5cb24b24cdf54660b5ed46324c035d41d97d8c0cb816723

See more details on using hashes here.

Provenance

The following attestation bundles were made for mnemoseed-0.1.0-py3-none-any.whl:

Publisher: release.yml on MnemoSeed/mnemoseed

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page