Skip to content
Palimem Docsspec v1.7.0

Release notes — v1.7.0

Palimem v1.7.0 introduced the current component spec. Local-first governed memory for coding agents — SQLite WAL, MCP stdio, supersession, scoped recall, and audit export.

Highlights

Section titled “Highlights”
  • 11 MCP tools — including memory_query_temporal and memory_audit_export
  • 143 black-box validation behaviors (spec v1.7.0)
  • Phase 6 integrations — Cursor, Copilot CLI, Codex, Windsurf, VS Code Copilot Agent, Copilot IDE, Gemini CLI
  • npx install path — @palimem/mcp on npm; connect MCP harnesses without cloning

Quick install (npx)

Section titled “Quick install (npx)”

From any project directory, with Node.js 18+:

Terminal window
npx @palimem/mcp ai-memory connect <harness> \
  --project-root "$(pwd)" \
  --launcher npx \
  --data-dir .ai-memory/data

Replace <harness> with one of: cursor, windsurf, copilot, codex, vscode, or gemini.

The --launcher npx flag writes MCP config with npx -y @palimem/mcp instead of a path to a local clone. See each integration page for harness-specific targets and flags.

Example (Cursor):

Terminal window
npx @palimem/mcp ai-memory connect cursor \
  --project-root "$(pwd)" \
  --launcher npx \
  --data-dir .ai-memory/data

Claude Code uses a copied .mcp.json or the bundled plugin — see Claude Code and Claude Code plugin.

Run the MCP server directly (no connect helper):

Terminal window
npx @palimem/mcp

npm package

Section titled “npm package”

@palimem/mcp is published on npm (current: v1.7.2). Install and connect without cloning the repository. PyPI packages are not yet published.

Verify

Section titled “Verify”

From a cloned repository (contributors and operators):

Terminal window
python3 tests/run_validation.py
bash integrations/run_readiness.sh
python3 benchmarks/run_benchmarks.py --strict

After connecting your harness, ask your agent to call tools/list (expect 11 tools) and memory_status (expect ok: true and index_status.state: "current").

Next steps

Section titled “Next steps”