PLUR vs Cognee

Choose Cognee for an open, self-hostable graph+vector pipeline. Choose PLUR for a portable engram format you own and share as packs.

Bottom line: Cognee if you want an ECL graph+vector memory pipeline; PLUR if you want to own the memory as an open, portable format with shareable knowledge packs.

What each is best at

• Cognee — an open, self-hostable ECL (extract-cognify-load) graph+vector memory pipeline, EU-funded, with an official cognee-mcp server. Its strength is a solid graph pipeline you can run yourself.
• PLUR — memory you own: a typed, open YAML engram format (published spec) that runs fully local, is shared across tools over MCP, and bundles into portable knowledge packs.

Where they differ (that matters)

• Both are open and self-hostable — that's not the differentiator. The difference is what the memory is.
• Portable format vs graph pipeline. PLUR's memory is a portable engram format (read, diff, move); Cognee is a graph+vector store, not an open interchange spec.
• Knowledge packs. PLUR bundles and shares team memory as files; Cognee has no pack primitive.
• Graph depth — Cognee leads. If a graph pipeline is the point, Cognee is stronger today; PLUR's graph is lighter (graph-DB on the roadmap).

Recall — table stakes, not the deciding factor

Both are competitive; quality is table stakes. PLUR reaches 97.6% R@5 on LongMemEval-S, fully local. The decision is the open portable format and packs, not a recall delta. (LongMemEval-S · n=500 · chunk · canonical-doc; R@5 = evidence in the top-5, not answer accuracy; measured on our own plur-bench harness, public with our paper.)

Choose Cognee if

you want an open, self-hostable graph+vector pipeline as your memory model.

Choose PLUR if

you want a portable engram format you own, shareable knowledge packs, and one memory across the tools you already use.

Install PLUR

npx @plur-ai/mcp init      # Claude Code / Cursor / Windsurf (any MCP client)
openclaw plugins install @plur-ai/claw && openclaw gateway --force   # OpenClaw
pip install plur-hermes    # Hermes Agent (Python)

Engrams are stored locally as files under ~/.plur/. Connect over MCP from Claude Code, Cursor, Windsurf, OpenClaw, or Hermes.

FAQ

Cognee vs PLUR — which should I use? Cognee for a graph+vector pipeline; PLUR for a portable, owned engram format with shareable packs.

Both are open-source and EU-aligned — what's the real difference? What the memory is: Cognee is a graph store; PLUR is an open portable format you can read, diff, move, and share as packs.