Connect agents

Read this if you’re connecting OMem to a specific agent and want the exact steps. For the conceptual overview, see let an AI agent use your memory; this page is the wiring reference.

Skill first, MCP as the fallback

OMem ships one skill — a small, agent-neutral wrapper that teaches an agent when to query your work memory and how to walk the results. There are two ways an agent picks it up:

• If your agent has a skill / plugin mechanism, use the skill. It’s the richer integration: the skill carries trigger semantics (when to auto-query) and the progressive-disclosure flow, not just raw tools. This is the preferred path for Claude Code, Codex, and Claw Hub.
• If your agent only speaks MCP, use the omem mcp server. Same six tools, but as a generic MCP endpoint — the fallback for agents without a native skill mechanism (Cursor, Cline, Continue, Zed, …).

Both surface the same six read-only capabilities; the difference is whether the agent also gets the skill’s trigger + drill guidance.

Agents with a skill

Claude Code

With the omem CLI installed and set up, add the marketplace and install the skill:

/plugin marketplace add seacen/omem
/plugin install omem@omem-marketplace

The agent now auto-invokes OMem on work-context questions. Verify by asking “check my OMem health” — it should run omem doctor.

Codex

OMem ships a Codex plugin that carries the same skill. With the omem CLI installed, install it from Codex’s plugin browser:

1. In Codex, run /plugins to open the plugin browser.
2. Add OMem’s source: seacen/omem (the GitHub repo). Use the marketplace tabs to switch to it.
3. Open the OMem plugin and choose Install plugin.

That’s all — the plugin brings OMem’s skill with it, so Codex now auto-queries your work memory the same way Claude Code does. Verify by asking “check my OMem health” — it should run omem doctor.

Codex's plugin browser is interactive

Codex installs third-party plugins through the interactive /plugins browser rather than a one-line command — you add the seacen/omem source and pick OMem from the list. The plugin carries the same skill content as the Claude Code one.

Claw Hub (OpenClaw)

OMem’s skill is published on Claw Hub, OpenClaw’s public skill registry. With the omem CLI installed, install the skill by its slug:

openclaw skills install omem

(Add --global to install it for every workspace.) You can also browse or search it at clawhub.ai, or use the standalone Claw Hub CLI:

npm i -g clawhub
clawhub install omem

One skill, three front doors

Claude Code, Codex, and Claw Hub all reach the same skill content — identical triggers, identical six tools. You’re not configuring three different things; you’re installing one skill through whichever door your agent uses.

Agents without a skill — the MCP server

omem mcp runs a stdio MCP server exposing the six tools to any MCP client. One config entry connects it:

Cursor — .cursor/mcp.json (or Settings → MCP Servers):

{
  "mcpServers": {
    "omem": { "command": "omem", "args": ["mcp"] }
  }
}

Cline, Continue, Zed — same shape (command: "omem", args: ["mcp"]) in that agent’s MCP config file.

When to use MCP if your agent also has a skill

If your agent supports a skill, prefer it — you get the trigger and drill guidance for free. Reach for the raw MCP entry only when the agent has no skill mechanism, or when you specifically want to wire the tools yourself.

The six tools (skill = MCP)

Both surfaces expose exactly these, and nothing else:

Tool	What it does
query	Ranked search across all indexed work
page_get	Full wiki page (frontmatter + body)
raw_get_path	Path to the original source file (file kind)
raw_get_parsed_path	Path to the parsed source (with version history)
wiki_ls	Browse pages by kind / source / account / date
doctor	OMem health check

Read-only by design

Setup, ingest, re-indexing, purging, config changes — none are exposed to agents, on either surface. They stay deliberate human actions, so connecting an agent can never let it reconfigure or delete your data. See querying from an agent.

Prerequisite (all paths)

Every integration needs the omem CLI installed and configured first:

curl -fsSL https://github.com/seacen/omem/releases/latest/download/install.sh | sh
omem setup
omem ingest

See install and setup.

What’s next

• Querying from an agent — how a well-behaved agent uses these tools.
• Schedule ingest — keep the memory fresh automatically.