Back to Gap Map Product / UX / Agent memory & context stores

memU

NevaMind AI
source available / Overall score: 1.4

memU gives coding agents one personal memory shared across sessions, agents and devices. A scheduled task slices each host's session logs into jobs, the agent distills them under memU's instructions into Markdown memory and skill files, and memU indexes the result and retrieves relevant files before later tasks. It ships adapters for Codex, Claude Code, Cursor, OpenClaw and Hermes Agent. NevaMind AI builds it.

Openness

3 medium confidence
3.0
license
memU-Modified-Apache-2.0 (LICENSE.txt is the Apache License 2.0 text with the section 2 copyright grant rewritten in MIT wording and an added notice-retention condition, and other clauses reworded)
source
public
core features withheld
no — the 340-path tree at 12d144c holds the memory service, storage backends, embedding providers and every host adapter with no ee/, enterprise/, commercial/ or proprietary/ path, and memU Cloud is a free hosted backend for the same three operations
commercial
memU Cloud hosted backend at memu.so

memU's whole memory pipeline, from the session-log adapters to the storage and retrieval service, is in the public repository and runs locally with the user's own storage and embedding provider. Its LICENSE file is the Apache 2.0 text with the copyright grant rewritten in MIT-style wording, so it is not the approved Apache license the project names elsewhere, though nothing in it limits who may use the software or at what scale.

Adoption

1 medium confidence
1.0

Adoption is read from monthly downloads of memu-cli on PyPI and its npm launcher, the package every memU host adapter installs whether it stores memory locally or in memU Cloud. The memu-py package from the project's earlier framework line is left out, and installs say nothing about how many agents keep using it.

Capability

2 medium confidence
2.0

memU retrieves across three separate stores of memory, workspace and skill files, and it supplies the instructions for distilling session logs into them. The distilling and merging are done by the host agent's model, not by memU, whose memory service makes no model calls of its own, so memU organizes and serves memory rather than deciding what to keep.

Verified 2026-11-20, except some scores, still awaiting re-confirmation