ZMem

Builders

Build tools, agents, and product workflows on top of ZMem.

ZMem is useful as a standalone memory tool, but the deeper value comes from building agent workflows that can remember, explain, and hand off state without hiding where beliefs came from.

Builder Surfaces

Use the smallest surface that fits your integration:

  • CLI: best for scripts, local workflows, demos, and CI checks.
  • MCP server: best for agents that need memory through a tool protocol.
  • Local SQLite store: best for product surfaces that need direct inspection or dashboards.
  • Receipts and bundles: best for audit, debugging, and sharing evidence.
  • Handoff artifacts: best for moving state across projects, machines, or agent sessions.
zmem init --with-policy --with-agent-prompt --with-mcp-config
zmem status --summary-only
zmem agent pack --summary-only
zmem --db .zerker/memory.sqlite mcp

When an agent takes a memory-influenced action, keep the receipt id. That id is the bridge between the user-facing answer and the proof path.

zmem why <action-id>
zmem verify <action-id>
zmem bundle <action-id> --out-dir .zerker/exports

Product Patterns

Agent Console

Show memory status, review queues, receipt counts, and recent handoffs. Users should not need to know the storage schema to understand what an agent can currently trust.

Workflow Handoff

Create handoff artifacts at the end of meaningful work so another agent can resume with approved memory and proof context instead of a pasted transcript.

zmem handoff --summary-only
zmem --db .zerker/imported.sqlite restore --handoff-dir .zerker/handoff

Proof View

For any important action, expose a compact explanation: what memory was retrieved, what was injected, what was withheld, and whether the receipt verifies locally.

Design Principle

Do not make users choose between convenience and trust. The best ZMem integrations make memory feel native to the agent while keeping provenance, review, and verification one click away.

On this page