memfor
Theme
Motion
git clone https://github.com/r5rana/agentware.git
cd agentware
git checkout 3c25cdd
claude "hi"
git clone https://github.com/r5rana/agentware.git
cd agentware
git checkout 3c25cdd
codex "hi"

This installs agentware, the open-source engine behind memfor. No tagged release yet. These commands pin commit 3c25cdd.

Apache-2.0. No memfor account. Needs macOS or Linux (Windows via WSL), git, Python 3, bash, jq, and Claude Code or Codex.

Send setup steps to my computer Join the Teams waitlist

Changelog. What shipped, with the commit that shipped it.

agentware has no tagged release yet, so entries point at commits on main. We write these by hand.

  1. First community pull request merged

    An outside contributor added the coding-interview practice skill (#10).

    Commit 3c25cdd

  2. Single-pass adversarial review

    Changes to agentware's own code and steering get one adversarial audit and a self-verified fix before they land.

    Commit 0c41fc9

  3. Approval-gated skill cache

    Skills run from a vetted, approval-gated cache, so a skill synced from a team repository doesn't run until it's approved.

    Commit c16e247

  4. Signed-tag updater

    The package updater only moves to signed release tags. It's off by default.

    Commit 34bc4b1

  5. Local models through LM Studio (experimental)

    Run the execute phase on a local model while planning stays on your cloud runtime.

    Commit a04d3a9

  6. Opt-in semantic recall and Codex support

    Mode B adds a pinned, on-device embedding model on top of BM25. It's opt-in, and BM25 stays the default. OpenAI Codex joins Claude Code as a supported runtime.

    Commits 938486c 08157eb

  7. Apache-2.0 licence added

    agentware gets its LICENSE and NOTICE files.

    Commit 60016ee

  8. LongMemEval-S scorer and baseline

    Recall@5 0.9140 on 470 answerable questions, recorded from a clean tree. Receipt

    Commit 42d58a0

  9. Repository created

    The first commit holds only a README.

    Commit 19834fb

What our status labels mean.

Available
on main, documented, covered by tests.
Opt-in
available but off by default.
Experimental
available, may change, known limits listed.
Next
built but not yet on main.
Early access · not built
a planned paid product. Waitlist only.
Exploring
an idea with no commitment.
Esc