openclaw-bee

Install and configure BEE — the Belief Extraction Engine for OpenClaw. Gives agents persistent structured memory across sessions. Auto-extracts beliefs at session end, scopes by agent namespace, and injects recalled context on agent start. Use when: setting up agent memory, configuring belief persistence, or troubleshooting BEE. NOT for: general memory questions (use built-in memory tools instead).

Safety Notice

This listing is from the official public ClawHub registry. Review SKILL.md and referenced scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "openclaw-bee" with this command: npx skills add vashkartik/openclaw-bee

BEE — Belief Extraction Engine

Give your OpenClaw agents persistent, structured memory across sessions.

What BEE Does

BEE hooks into the OpenClaw lifecycle and:

  • Extracts beliefs at session end via a lightweight LLM call (Haiku by default)
  • Injects recalled context before every agent spawn
  • Scopes by namespace — each agent (VECTOR, FORGE, ORACLE, etc.) has isolated beliefs
  • Deduplicates — cosine similarity check prevents duplicate beliefs (>0.92 → merge)
  • Tracks spawns — monitors subagent budget per session

Beliefs live in a SQLite database (vector.db) and persist indefinitely across restarts.


Installation

Step 1 — Install the package

From npm (recommended):

npm install -g @skysphere-labs/openclaw-bee

From GitHub (latest):

npm install -g github:skysphere-labs/openclaw-bee

Step 2 — Configure openclaw.json

Add BEE to your extensions in ~/.openclaw/openclaw.json:

{
  "extensions": {
    "entries": {
      "bee": {
        "enabled": true,
        "config": {
          "dbPath": "~/.openclaw/workspace/state/vector.db",
          "agentId": "main",
          "extractionEnabled": true,
          "extractionModel": "anthropic/claude-haiku-4-5",
          "maxCoreBeliefs": 10,
          "maxActiveBeliefs": 5,
          "maxRecalledBeliefs": 5
        }
      }
    }
  }
}

Step 3 — Restart the gateway

openclaw gateway restart

BEE will run its database migration on first start and begin capturing beliefs.


Configuration Options

OptionDefaultDescription
dbPathrequiredPath to your SQLite database
agentId"main"Namespace for belief scoping
extractionEnabledtrueEnable/disable belief extraction
extractionModel"anthropic/claude-haiku-4-5"Model used for extraction (cheapest works well)
extractionMinConfidence0.55Minimum confidence to store a belief (0-1)
maxCoreBeliefs10Core beliefs injected into every session
maxActiveBeliefs5Recently active beliefs injected
maxRecalledBeliefs5Semantically recalled beliefs per query
maxOutputChars2000Max chars of belief context injected
debugfalseEnable verbose logging
spawnBudgetWarning20Warn when subagent spawns exceed this threshold

Verifying It Works

After restart, ask your agent:

"How many beliefs do you have?"

Or check directly:

sqlite3 ~/.openclaw/workspace/state/vector.db "SELECT COUNT(*) FROM beliefs"

You should see beliefs accumulate after sessions complete.


Multi-Agent Setup

For setups with multiple named agents (VECTOR, FORGE, ORACLE, etc.), use different agentId values per agent spawn. BEE scopes beliefs by agentId so each PM has isolated memory.


Source

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Related Skills

Related by shared tags or category signals.

Automation

Xiaohongshu Ops

小红书端到端运营:账号定位、选题研究、内容生产、发布执行、数据复盘。 Use when: (1) 用户要写小红书笔记/帖子, (2) 用户说"发小红书"/"写个种草文"/"出一篇小红书", (3) 用户讨论小红书选题/热点/爆款分析/竞品对标, (4) 用户提到账号定位/人设/内容方向规划, (5) 用户要求生成...

Registry SourceRecently Updated
Automation

WeMP Ops

微信公众号全流程运营:选题→采集→写作→排版→发布→数据分析→评论管理。 Use when: (1) 用户要写公众号文章或提供了选题方向, (2) 用户说"写一篇关于XXX的文章"/"帮我写篇推文"/"出一篇稿子", (3) 用户要求采集热点/素材/竞品分析, (4) 用户提到公众号日报/周报/数据分析/阅读量/...

Registry SourceRecently Updated
Automation

agent-stock

用于股票行情查询与分析的命令行技能。用户提到 stock 命令、股票代码、最新资讯、市场概览、K 线或配置管理时调用。

Registry SourceRecently Updated