KMP — Agent memory that remembers why

Local first. Evidence attached. Time included.

KMP gives Codex and Claude Code local-first memory that preserves what happened, when and why. It stores decisions and evidence, not transcripts, on embedded SQLite, and exposes them through eleven memory tools plus three semantic view tools over a shared ChronoLoom view.

Ask “Show me the memory behind this decision.” The agent retrieves the evidence, opens ChronoLoom at the relevant moment and lights up the proof path. ChronoLoom names who is guiding, shows why the agent moved, and lets you take control without losing the frame or its undo. A shared time window crosses real 3D about layers; a flat camera shows the same memory, with labels as filters.

Why KMP?

Agents are good at doing the work in front of them. Tomorrow is harder. A decision survives, its rationale disappears, and somebody gets to rediscover the same incident from scratch.

KMP gives the agent a typed, temporal memory instead of a bag of text:

• evidence decides what can be claimed;
• relations carry the reason two memories belong together;
• time travel is explicit, cursor-based and auditable;
• old decisions are superseded, never quietly rewritten;
• UNKNOWN is an honest answer when the evidence is not there.

The normal path runs entirely on your machine. No KMP account. No hosted memory server. No Underpass cloud.

Install. Run setup. Done.

Codex CLI

codex plugin marketplace add underpass-ai/kmp --ref marketplace
codex plugin add kmp@underpass

Then ask Codex to run kmp-setup and restart Codex once.

Claude Code

/plugin marketplace add underpass-ai/kmp@marketplace
/plugin install kmp@underpass
/kmp:setup

Restart Claude Code once. Verify either host with its kmp-doctor workflow, or from a terminal:

kmp-mcp info
kmp-mcp doctor

That is the happy path. The plugin owns the MCP registration, so do not add a second KMP server by hand. Store selection and repair live in Embedded KMP.

Talk to it like a human

You normally ask for the outcome. The KMP skill chooses the memory moves.

KMP is memory, not surveillance. Store durable decisions and evidence, not transcripts.

It also waits to be asked. A session that never mentions memory makes no KMP call at all — the agent works from what is in front of it. Naming KMP, running a /kmp:* command, or opting in from your project's CLAUDE.md or AGENTS.md is what opens a route. If you would rather it enter known work on its own:

kmp-mcp config memory-routing always
kmp-mcp config memory-routing on-request   # the default

How it works — the 10-second version

flowchart LR
    U[You] --> H[Codex or Claude]
    H --> S[KMP skill]
    S -->|chooses a move| M[kmp-mcp]
    M --> K[embedded kernel]
    K --> D[(.kernel/\nSQLite)]
    K --> V[local read-only viewer]

The plugin installs the skills and declares one local MCP process. The skill turns intent into one or more of fifteen typed tools. kmp-mcp validates the request, and the kernel reads or writes the local graph-temporal store. The agent—not KMP—turns returned evidence into conversational prose.

ChronoLoom — memory you can see

Ask your agent: “Show me the memory behind this decision.” ChronoLoom opens on the evidence and lights up its proof path.

You control the same view: your agent can steer it; you can click, filter, pan or undo at any time.

Explore ChronoLoom · Technical architecture

Who owns what?

Human workflows such as kmp-setup, kmp-doctor, kmp-info, kmp-catchup, kmp-save, kmp-restore and kmp-revert compose the MCP surface. They are not extra memory verbs. The machine-checked ownership map is plugins/kmp/capabilities.json.

Local means local

.kernel/ is machine state and is ignored by git. A project-scoped store also maintains .kmp/memory.jsonl; it leaves your machine only if you deliberately commit or copy it.

Language without flattening the evidence