openclaw-memory-docs

OpenClaw plugin: Documentation Memory.

A conservative, audit-friendly memory store for project documentation and long-lived notes.

• No automatic capture - only explicit /remember-doc commands store data
• Local JSONL store with local deterministic embeddings (no external services)
• Optional secret redaction (API keys, tokens, private key blocks)
• Tag and project metadata for organization and filtering

Install

ClawHub

clawhub install openclaw-memory-docs

Dev

openclaw plugins install -l ~/.openclaw/workspace/openclaw-memory-docs
openclaw gateway restart

Commands

/remember-doc - Save a memory

/remember-doc [--tags t1,t2] [--project name] <text>

Stores a documentation memory item. If secrets are detected, they are redacted before storage.

Flags:

| Flag | Description | |------|-------------| | --tags t1,t2 | Comma-separated tags (merged with defaultTags from config) | | --project name | Associate this item with a project name |

Examples:

/remember-doc Dubai: decide A vs B, then collect facts, then prepare a tax advisor briefing.
/remember-doc --tags legal,tax --project dubai Tax advisor meeting scheduled for March.
/remember-doc --tags=api --project=backend The /users endpoint requires Bearer auth.

/search-docs - Search memories

/search-docs [--tags t1,t2] [--project name] <query> [limit]

Searches stored memories using semantic-ish similarity. Returns scored results with IDs, tags, and project badges.

Parameters:

| Parameter | Description | Default | |-----------|-------------|---------| | query | Search query text (required) | - | | limit | Max results to return (1-20) | 5 | | --tags t1,t2 | Filter results to items matching these tags | - | | --project name | Filter results to items with this project | - |

Examples:

/search-docs Dubai plan
/search-docs --project=dubai tax advisor 10
/search-docs --tags=api,backend endpoint auth

/list-docs - List recent memories

/list-docs [--tags t1,t2] [--project name] [limit]

Lists the most recent documentation memory items, with IDs, dates, tags, and project badges.

Parameters:

| Parameter | Description | Default | |-----------|-------------|---------| | limit | Max items to return (1-50) | 10 | | --tags t1,t2 | Filter to items matching these tags | - | | --project name | Filter to items with this project | - |

Examples:

/list-docs
/list-docs 20
/list-docs --project=dubai
/list-docs --tags=legal --project=dubai 5

/forget-doc - Delete a memory

/forget-doc <id>

Deletes a documentation memory item by its ID. Use /list-docs to find item IDs. Requires auth.

Example:

/forget-doc a1b2c3d4-e5f6-7890-abcd-ef1234567890

/export-docs - Export as markdown files

/export-docs [--tags t1,t2] [--project name] [path]

Exports documentation memories as individual markdown files for git-first workflows. Each file gets YAML frontmatter (id, kind, createdAt, tags, project) and the memory text as body.

Parameters:

| Parameter | Description | Default | |-----------|-------------|---------| | path | Target directory for exported files | exportPath config or ~/.openclaw/workspace/memory/docs-export | | --tags t1,t2 | Export only items matching these tags | - | | --project name | Export only items with this project | - |

File naming: YYYY-MM-DD_<shortid>.md (e.g., 2026-01-15_abc12345.md)

Examples:

/export-docs
/export-docs ~/docs/memories
/export-docs --project=dubai
/export-docs --tags=api --project=backend ~/exports

/import-docs - Import from markdown files

/import-docs [path]

Imports documentation memories from a directory of exported markdown files. Each .md file must have YAML frontmatter with id, kind, and createdAt fields (the format produced by /export-docs). Requires auth.

Duplicate items (matching by ID) and invalid files are skipped automatically.

Parameters:

| Parameter | Description | Default | |-----------|-------------|---------| | path | Source directory containing markdown files | exportPath config or ~/.openclaw/workspace/memory/docs-export |

Examples:

/import-docs
/import-docs ~/docs/memories
/import-docs /path/to/exported/docs

Tool

docs_memory_search

Available to agents and automations as a tool call.

{
  "query": "Dubai plan A vs B",
  "limit": 5,
  "tags": ["legal"],
  "project": "dubai"
}

Parameters:

| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | query | string | yes | Search query text | | limit | number | no | Max results (1-20, default 5) | | tags | string[] | no | Filter results to items matching all given tags | | project | string | no | Filter results to items with this project name |

Returns a hits array with score, id, createdAt, tags, project, and text for each match.

Config

{
  "plugins": {
    "entries": {
      "openclaw-memory-docs": {
        "enabled": true,
        "config": {
          "storePath": "~/.openclaw/workspace/memory/docs-memory.jsonl",
          "dims": 256,
          "redactSecrets": true,
          "defaultTags": ["docs"],
          "maxItems": 5000,
          "exportPath": "~/.openclaw/workspace/memory/docs-export"
        }
      }
    }
  }
}

Options:

| Option | Type | Default | Description | |--------|------|---------|-------------| | enabled | boolean | true | Enable or disable the plugin | | storePath | string | ~/.openclaw/workspace/memory/docs-memory.jsonl | Path to the JSONL storage file | | dims | number | 256 | Embedding dimensions (32-2048) | | redactSecrets | boolean | true | Redact detected secrets before storage | | defaultTags | string[] | ["docs"] | Tags automatically added to every saved item | | maxItems | number | 5000 | Maximum items in the store (100-100000) | | exportPath | string | ~/.openclaw/workspace/memory/docs-export | Default directory for /export-docs and /import-docs |

Design Notes

• This plugin intentionally does not auto-capture messages.
• If you want automatic capture, use openclaw-memory-brain instead.
• Secret redaction covers common patterns: API keys, tokens, private key blocks.
• Embeddings are deterministic and local - no calls to external embedding services.
• Export/import uses a git-friendly format: one markdown file per memory item, with deterministic filenames (YYYY-MM-DD_<shortid>.md). Frontmatter contains all metadata, making diffs clean and merge-friendly.