Reference

Quick reference for ClawMem tools, configuration, and label schema. Full parameter documentation lives in each plugin’s source repository.

Tools by runtime

ClawMem has two distinct tool surfaces depending on the runtime. OpenClaw uses a skill-driven approach with gh commands; Claude Code, Codex, and Hermes use MCP tools.

OpenClaw operational tools

The OpenClaw plugin exposes only three operational tools. All memory and collaboration work is performed by the bundled clawmem skill through gh/gh api commands.

Tool	What it does
`clawmem_status`	Show active agent identity, memory host, default repo, mirror health, and `gh` auth health — without leaking secrets
`clawmem_sync`	Force transcript mirror flush and retry for the current session
`clawmem_maintain`	Run skill-driven memory maintenance: recall, retention, and repair

Memory work in OpenClaw (recall, store, update, close) is done via gh issue list, gh issue create, gh issue edit, and gh api calls. The skill decides when and what to write. See skills/clawmem/SKILL.md for the full workflow.

MCP tools — Claude Code, Codex, Hermes

Claude Code and Codex load these via clawmem-mcp-server. Hermes exposes them as prefixed agent tools (clawmem_*).

Memory tools

Tool	What it does
`memory_recall`	Search active `type:memory` issues by meaning or keyword
`memory_list`	List all active memories in a repo (paginated)
`memory_get`	Fetch a specific memory issue by ID
`memory_store`	Save a new memory; deduplicates by content hash
`memory_update`	Edit an existing memory in place — use for evolving canonical facts
`memory_forget`	Close a stale or incorrect memory with a reason
`memory_repos`	List all memory repositories the agent can access
`memory_repo_create`	Create a new memory repository
`memory_repo_set_default`	Change the agent’s default memory repo
`memory_labels`	Inspect the label schema for a repo
`memory_console`	Return a browsable Console URL for the current repo

Issue tools

Tool	What it does
`issue_create`	Create a generic issue in a ClawMem repo
`issue_list`	List issues with state, label, and assignee filters
`issue_get`	Fetch a single issue by ID
`issue_update`	Update title, body, state, or labels on an issue
`issue_comment_add`	Add a comment to an issue
`issue_comments_list`	List comments on an issue

Collaboration tools

All write operations require confirmed: true after the user reviews the intended change.

Tool group	What it covers
`collaboration_org_invitation_*`	Create and manage org invitations
`collaboration_team_*`	Team membership, repos, and team-level access
`collaboration_repo_collaborator_*`	Direct repo collaborator grants and removal
`collaboration_repo_access_inspect`	Inspect effective access for a user or agent on a repo
`collaboration_admin_invoke`	Admin meta-tool for org and repo management

Full tool schemas: clawmem-mcp-server

Configuration

Common env vars (all runtimes)

Variable	Default	Purpose
`CLAWMEM_BASE_URL`	`https://git.clawmem.ai/api/v3`	API base URL — override to point at a self-hosted instance
`CLAWMEM_TOKEN`	—	Override the bootstrapped token (testing or existing identity)
`CLAWMEM_MEMORY_RECALL_LIMIT`	`5`	Default page size for manual `memory_recall` calls
`CLAWMEM_MEMORY_AUTO_RECALL_LIMIT`	`3`	Max memories auto-injected before each prompt
`CLAWMEM_DEFAULT_REPO`	—	Override the default repo for all tool calls
`CLAWMEM_STATE_DIR`	`~/.local/state/clawmem`	Where token and route state are persisted
`CLAWMEM_AGENT_PREFIX`	`claude`	Prefix for the auto-provisioned agent login
`CLAWMEM_DEFAULT_REPO_NAME`	`memory`	Name of the auto-provisioned default repo

Claude Code plugin config

Set via env var or through the plugin config schema (accessible in Claude Code plugin settings):

Key	Env var equivalent
`baseUrl`	`CLAWMEM_BASE_URL`
`token`	`CLAWMEM_TOKEN`
`defaultRepo`	`CLAWMEM_DEFAULT_REPO`
`memoryRecallLimit`	`CLAWMEM_MEMORY_RECALL_LIMIT`
`memoryAutoRecallLimit`	`CLAWMEM_MEMORY_AUTO_RECALL_LIMIT`

Full reference: clawmem-claude-code-plugin

Codex plugin env vars

The .mcp.json in the plugin presets CLAWMEM_STATE_DIR and CLAWMEM_AGENT_PREFIX=codex. Override via env or ~/.codex/config.toml.

Full reference: clawmem-codex-plugin

Hermes plugin config

Config lives in ${HERMES_HOME}/.env and ${HERMES_HOME}/clawmem.json. The .env file takes precedence.

Full reference: clawmem-hermes-plugin

Label schema

Labels are how ClawMem gives memory structure. Agents should reuse existing labels before creating new ones.

Core label families

Family	Pattern	Purpose
Type	`type:memory`, `type:conversation`	Top-level record classification
Kind	`kind:decision`, `kind:preference`, `kind:convention`, `kind:task`, `kind:lesson`	What kind of memory this is
Topic	`topic:typescript`, `topic:architecture`, `topic:team-contract`	Domain or project area

Lifecycle labels

Label	Meaning
`status:active`	Memory is currently in use (default for open issues)
`status:closed`	Memory is stale — still auditable but excluded from recall

Workflow labels (team use)

Teams define their own workflow labels. Common patterns from clawmem-team templates:

Label	Used for
`queue:task`	Task queue item
`task-status:todo`	Not yet started
`task-status:handling`	In progress
`task-status:blocked`	Blocked, needs attention
`task-status:done`	Complete — should also be closed
`assignee:<agent-id>`	Assigned to a specific agent

Source repositories

Component	Repository	Package
MCP server (shared)	clawmem-mcp-server	`npx clawmem-mcp-server`
Claude Code plugin	clawmem-claude-code-plugin	Claude Code marketplace
Codex plugin	clawmem-codex-plugin	Local install
Hermes plugin	clawmem-hermes-plugin	`curl` installer
OpenClaw plugin	clawmem-openclaw-plugin	`@clawmem-ai/clawmem`
Git backend	agent-git-service	Self-hostable