SuperLocalMemory works with any MCP client through slm connect <ide>, which
writes the MCP server into that tool's config. A plugin adds more on top: skills
that teach the agent when to recall and what to save, sub-agents, session hooks
and slash commands. Every plugin is built from one source (plugin-src/), so
they all carry the same 15 skills and 4 sub-agents, adapted to each host.
Install SuperLocalMemory first, from npm or PyPI (getting started).
The npm package also carries every plugin folder, at
$(npm root -g)/superlocalmemory/; with a PyPI install, the same folders are in
this repository.
| Host | Install | What you get |
|---|---|---|
| Claude Code | claude plugin marketplace add qualixar/superlocalmemory then claude plugin install superlocalmemory@qualixar |
MCP server, 15 skills, 4 sub-agents, hooks, /slm-loop |
| Codex | slm connect codex then slm codex install |
MCP server, skills, sub-agents, lifecycle hooks |
| VS Code / GitHub Copilot | slm connect vscode-copilot --here (from the project root), then copy copilot-plugin/.github/ add-ons |
MCP server, agent rules, 15 prompt files, 4 agents, hooks |
| Cursor-format hosts (Grok Bot, Cursor) | Add the qualixar marketplace in the host's Plugins screen, then superlocalmemory |
MCP server, a bot-sized skill set, no hooks or dashboard needed |
| Hermes | hermes plugins install 'https://github.com/qualixar/superlocalmemory.git#hermes-plugin' |
Skills, child-agent roles, /slm commands |
Antigravity (agy) |
agy plugin install "$(npm root -g)/superlocalmemory/antigravity-plugin" |
MCP server, skills, agents, hooks |
| Any other agent | Paste the universal agent rules into AGENTS.md, CLAUDE.md, .cursorrules or the system prompt |
Memory discipline for any MCP client |
| Web apps (ChatGPT, Claude on the web, Muse, Composio) | Web access in the dashboard, then the host guides and web agent instructions | Recall and, if allowed, save through Web access |
| Skill | What it covers |
|---|---|
slm-recall |
Recall and search: filters, tags, kinds, time windows, the answer check |
slm-remember |
What to save, memory kinds, replaces, corrections |
slm-session |
Session start, pinned items and standing rules, closing a session |
slm-status |
Health, slm db integrity and slm db repair, embedder and model checks |
slm-governance |
Scopes, roles, retention, audit trail, GDPR export and erasure |
slm-scope |
Personal, shared and global memory across profiles |
slm-profile |
Memory profiles and switching between them |
slm-graph |
The knowledge graph and entity tools |
slm-mesh |
SLM-Mesh: messages, locks and shared state between agent sessions |
slm-loop |
Bounded loops and their history |
slm-cache, slm-compress |
Context optimization: KV cache and reversible compression |
slm-web-access |
Setting up and diagnosing Web access for web apps |
slm-bot-memory, slm-getting-started-bot |
Several bots sharing one computer and one store |
The four sub-agents are the memory advisor, the optimize advisor, the governance advisor and the loop runner.
claude plugin marketplace add qualixar/superlocalmemory
claude plugin install superlocalmemory@qualixarslm connect claude-code prints the same two commands. Updates arrive through
claude plugin marketplace update qualixar. Manual MCP setup without the plugin
is in IDE setup.
slm connect codex --dry-run # show what would be written
slm connect codex # add the MCP server to your Codex config
slm codex install # skills, sub-agents and lifecycle hooks (--dry-run to preview)
slm codex status # what is installed
slm codex remove # remove only what SLM installedCodex also installs it as a plugin from the same marketplace:
codex plugin marketplace add qualixar/superlocalmemory --ref main
codex plugin add superlocalmemory-codex@qualixarReview and trust the new hooks in Codex with /hooks. The agent rules in
codex-plugin/AGENTS.md are not written for you: append them to your
AGENTS.md (details).
From the project root:
slm connect vscode-copilot --hereThis merges the MCP server into .vscode/mcp.json and the agent rules into
.github/copilot-instructions.md, without overwriting what is there. The prompt
files, agents and hooks are in copilot-plugin/.github/: copy them into your
project's .github/ and keep any files of your own
(details).
Add the qualixar marketplace (.cursor-plugin/marketplace.json in this
repository) in the host's Plugins screen, then add superlocalmemory. The MCP
server starts with uvx pinned to the release, with no API key or sign-in. See
the README section
for what bots on one computer share. This plugin keeps its own memory on the Grok Bot
computer. To reach the memory on your own computer, add Web access as a custom MCP
server in Grok Bot's Plugins screen (steps).
hermes plugins install 'https://github.com/qualixar/superlocalmemory.git#hermes-plugin'To pin an exact release, pass its 40-character commit with --ref. Grant the
superlocalmemory MCP server to the plugin when Hermes asks. See
Hermes for settings and what the plugin records.
agy plugin install takes a local folder only. With an npm install:
agy plugin install "$(npm root -g)/superlocalmemory/antigravity-plugin"With a PyPI install, run scripts/install-antigravity-plugin.sh from this
repository: it fetches only antigravity-plugin/ into a local cache and installs
it from there. Safe to re-run.
slm upgrade-hosts # preview which installed hosts are behind
slm upgrade-hosts --host codex --apply
slm doctor # warns when a plugin is older than SLM