claude-mem/docs/public/openclaw-integration.mdx

---
title: OpenClaw Integration
description: Persistent memory for OpenClaw agents — observation recording, system prompt context injection, and real-time observation feeds
icon: dragon
---

## Overview

The OpenClaw plugin gives claude-mem persistent memory to agents running on the [OpenClaw](https://openclaw.ai) gateway. It handles three things:

1. **Observation recording** — Captures tool usage from OpenClaw's embedded runner and sends it to the claude-mem worker for AI processing
2. **System prompt context injection** — Injects the observation timeline into each agent's system prompt via the `before_prompt_build` hook, keeping `MEMORY.md` free for agent-curated memory
3. **Observation feed** — Streams new observations to messaging channels (Telegram, Discord, Slack, etc.) in real-time via SSE

<Info>
OpenClaw's embedded runner (`pi-embedded`) calls the Anthropic API directly without spawning a `claude` process, so claude-mem's standard hooks never fire. This plugin bridges that gap by using OpenClaw's event system to capture the same data.
</Info>

## How It Works

```plaintext
OpenClaw Gateway
  │
  ├── before_agent_start ───→ Init session
  ├── before_prompt_build ──→ Inject context into system prompt
  ├── tool_result_persist ──→ Record observation
  ├── agent_end ────────────→ Summarize + Complete session
  └── gateway_start ────────→ Reset session tracking + context cache
                    │
                    ▼
         Claude-Mem Worker (localhost:37777)
           ├── POST /api/sessions/init
           ├── POST /api/sessions/observations
           ├── POST /api/sessions/summarize
           ├── POST /api/sessions/complete
           ├── GET  /api/context/inject ──→ System prompt context
           └── GET  /stream ─────────────→ SSE → Messaging channels
```

### Event Lifecycle

<Steps>
  <Step title="Agent starts (before_agent_start)">
    When an OpenClaw agent starts, the plugin initializes a session by sending the user prompt to `POST /api/sessions/init` so the worker can create a new session and start processing.
  </Step>
  <Step title="Context injected (before_prompt_build)">
    Before each LLM call, the plugin fetches the observation timeline from the worker's `/api/context/inject` endpoint and returns it as `appendSystemContext`. This injects cross-session context directly into the agent's system prompt without writing any files.

    The context is cached for 60 seconds to avoid re-fetching on every LLM turn within a session.
  </Step>
  <Step title="Tool use recorded (tool_result_persist)">
    Every time the agent uses a tool (Read, Write, Bash, etc.), the plugin sends the observation to `POST /api/sessions/observations` with the tool name, input, and truncated response (max 1000 chars). This is fire-and-forget — it doesn't block the agent from continuing work.

    Tools prefixed with `memory_` are skipped to avoid recursive recording.
  </Step>
  <Step title="Agent finishes (agent_end)">
    When the agent completes, the plugin extracts the last assistant message and sends it to `POST /api/sessions/summarize`, then calls `POST /api/sessions/complete` to close the session. Both are fire-and-forget.
  </Step>
  <Step title="Gateway restarts (gateway_start)">
    Clears all session tracking (session IDs, context cache) so agents get fresh state after a gateway restart.
  </Step>
</Steps>

### System Prompt Context Injection

The plugin injects cross-session observation context into each agent's system prompt via OpenClaw's `before_prompt_build` hook. The content comes from the worker's `GET /api/context/inject?projects=<project>` endpoint, which generates a formatted markdown timeline from the SQLite database.

This approach keeps `MEMORY.md` under the agent's control for curated long-term memory (decisions, preferences, durable facts), while the observation timeline is delivered through the system prompt where it belongs.

<Info>
Context is cached for 60 seconds per project to avoid re-fetching on every LLM turn. The cache is cleared on gateway restart. Use `syncMemoryFileExclude` to opt specific agents out of context injection entirely.
</Info>

### Observation Feed (SSE → Messaging)

The plugin runs a background service that connects to the worker's SSE stream (`GET /stream`) and forwards `new_observation` events to a configured messaging channel. This lets you monitor what your agents are learning in real-time from Telegram, Discord, Slack, or any supported OpenClaw channel.

The SSE connection uses exponential backoff (1s → 30s) for automatic reconnection.

## Setting Up the Observation Feed

The observation feed sends a formatted message to your OpenClaw channel every time claude-mem creates a new observation. Each message includes the observation title and subtitle so you can follow along as your agents work.

Messages look like this in your channel:

```
🧠 Claude-Mem Observation
**Implemented retry logic for API client**
Added exponential backoff with configurable max retries to handle transient failures
```

### Step 1: Choose your channel

The observation feed works with any channel that your OpenClaw gateway has configured. You need two pieces of information:

- **Channel type** — The name of the channel plugin registered with OpenClaw (e.g., `telegram`, `discord`, `slack`, `signal`, `whatsapp`, `line`)
- **Target ID** — The chat ID, channel ID, or user ID where messages should be sent

<AccordionGroup>
  <Accordion title="Telegram" icon="telegram">
    **Channel type:** `telegram`

    **Target ID:** Your Telegram chat ID (numeric). To find it:
    1. Message [@userinfobot](https://t.me/userinfobot) on Telegram
    2. It will reply with your chat ID (e.g., `123456789`)
    3. For group chats, the ID is negative (e.g., `-1001234567890`)

    ```json
    "observationFeed": {
      "enabled": true,
      "channel": "telegram",
      "to": "123456789"
    }
    ```
  </Accordion>

  <Accordion title="Discord" icon="discord">
    **Channel type:** `discord`

    **Target ID:** The Discord channel ID. To find it:
    1. Enable Developer Mode in Discord (Settings → Advanced → Developer Mode)
    2. Right-click the channel → Copy Channel ID

    ```json
    "observationFeed": {
      "enabled": true,
      "channel": "discord",
      "to": "1234567890123456789"
    }
    ```
  </Accordion>

  <Accordion title="Slack" icon="slack">
    **Channel type:** `slack`

    **Target ID:** The Slack channel ID (not the channel name). To find it:
    1. Open the channel in Slack
    2. Click the channel name at the top
    3. Scroll to the bottom of the channel details — the ID looks like `C01ABC2DEFG`

    ```json
    "observationFeed": {
      "enabled": true,
      "channel": "slack",
      "to": "C01ABC2DEFG"
    }
    ```
  </Accordion>

  <Accordion title="Signal" icon="signal-messenger">
    **Channel type:** `signal`

    **Target ID:** The Signal phone number or group ID configured in your OpenClaw gateway.

    ```json
    "observationFeed": {
      "enabled": true,
      "channel": "signal",
      "to": "+1234567890"
    }
    ```
  </Accordion>

  <Accordion title="WhatsApp" icon="whatsapp">
    **Channel type:** `whatsapp`

    **Target ID:** The WhatsApp phone number or group JID configured in your OpenClaw gateway.

    ```json
    "observationFeed": {
      "enabled": true,
      "channel": "whatsapp",
      "to": "+1234567890"
    }
    ```
  </Accordion>

  <Accordion title="LINE" icon="line">
    **Channel type:** `line`

    **Target ID:** The LINE user ID or group ID from the LINE Developer Console.

    ```json
    "observationFeed": {
      "enabled": true,
      "channel": "line",
      "to": "U1234567890abcdef"
    }
    ```
  </Accordion>
</AccordionGroup>

### Step 2: Add the config to your gateway

Add the `observationFeed` block to your claude-mem plugin config in your OpenClaw gateway configuration:

```json
{
  "plugins": {
    "claude-mem": {
      "enabled": true,
      "config": {
        "project": "my-project",
        "observationFeed": {
          "enabled": true,
          "channel": "telegram",
          "to": "123456789"
        }
      }
    }
  }
}
```

<Warning>
The `channel` value must match a channel plugin that is already configured and running on your OpenClaw gateway. If the channel isn't registered, you'll see `Unknown channel type: <channel>` in the logs.
</Warning>

### Step 3: Verify the connection

After starting the gateway, check that the feed is connected:

1. **Check the logs** — You should see:
   ```
   [claude-mem] Observation feed starting — channel: telegram, target: 123456789
   [claude-mem] Connecting to SSE stream at http://localhost:37777/stream
   [claude-mem] Connected to SSE stream
   ```

2. **Use the status command** — Run `/claude_mem_feed` in any OpenClaw chat to see:
   ```
   Claude-Mem Observation Feed
   Enabled: yes
   Channel: telegram
   Target: 123456789
   Connection: connected
   ```

3. **Trigger a test** — Have an agent do some work. When the worker processes the tool usage into an observation, you'll receive a message in your configured channel.

<Info>
The feed only sends `new_observation` events — not raw tool usage. Observations are generated asynchronously by the worker's AI agent, so there's a 1-2 second delay between tool use and the observation message appearing in your channel.
</Info>

### Troubleshooting the Feed

| Symptom | Cause | Fix |
|---------|-------|-----|
| `Connection: disconnected` | Worker not running or wrong port | Check `workerPort` config, run `npm run worker:status` |
| `Connection: reconnecting` | Worker was running but connection dropped | The plugin auto-reconnects with backoff — wait up to 30s |
| `Unknown channel type` in logs | Channel plugin not loaded on gateway | Verify your OpenClaw gateway has the channel plugin configured |
| No messages appearing | Feed connected but no observations being created | Check that agents are running and the worker is processing observations |
| `Observation feed disabled` in logs | `enabled` is `false` or missing | Set `observationFeed.enabled` to `true` |
| `Observation feed misconfigured` in logs | Missing `channel` or `to` | Both `channel` and `to` are required |

## Installation

Run this one-liner to install everything automatically:

```bash
curl -fsSL https://install.cmem.ai/openclaw.sh | bash
```

The installer handles dependency checks (Bun, uv), plugin installation, memory slot configuration, AI provider setup, worker startup, and optional observation feed configuration.

You can also pre-select options:

```bash
# With a specific AI provider
curl -fsSL https://install.cmem.ai/openclaw.sh | bash -s -- --provider=gemini --api-key=YOUR_KEY

# Fully unattended (defaults to Claude Max Plan)
curl -fsSL https://install.cmem.ai/openclaw.sh | bash -s -- --non-interactive

# Upgrade existing installation
curl -fsSL https://install.cmem.ai/openclaw.sh | bash -s -- --upgrade
```

### Manual Configuration

Add `claude-mem` to your OpenClaw gateway's plugin configuration:

```json
{
  "plugins": {
    "claude-mem": {
      "enabled": true,
      "config": {
        "project": "my-project",
        "syncMemoryFile": true,
        "workerPort": 37777,
        "observationFeed": {
          "enabled": true,
          "channel": "telegram",
          "to": "your-chat-id"
        }
      }
    }
  }
}
```

<Note>
The claude-mem worker service must be running on the same machine as the OpenClaw gateway. The plugin communicates with it via HTTP on `localhost:37777`.
</Note>

## Configuration

<ParamField body="project" type="string" default="openclaw">
  Project name for scoping observations in the memory database. All observations from this gateway will be stored under this project name.
</ParamField>

<ParamField body="syncMemoryFile" type="boolean" default={true}>
  Inject observation context into the agent system prompt via `before_prompt_build` hook. When `true`, agents receive cross-session context automatically. Set to `false` to disable context injection entirely (observations are still recorded).
</ParamField>

<ParamField body="syncMemoryFileExclude" type="string[]" default={[]}>
  Agent IDs excluded from automatic context injection. Useful for agents that curate their own memory and don't need the observation timeline (e.g., `["snarf", "debugger"]`). Observations are still recorded for excluded agents — only the context injection is skipped.
</ParamField>

<ParamField body="workerPort" type="number" default={37777}>
  Port for the claude-mem worker service. Override if your worker runs on a non-default port.
</ParamField>

<ParamField body="observationFeed.enabled" type="boolean" default={false}>
  Enable live observation streaming to messaging channels.
</ParamField>

<ParamField body="observationFeed.channel" type="string">
  Channel type: `telegram`, `discord`, `signal`, `slack`, `whatsapp`, `line`
</ParamField>

<ParamField body="observationFeed.to" type="string">
  Target chat/user/channel ID to send observations to.
</ParamField>

## Commands

### /claude_mem_feed

Show or toggle the observation feed status.

```
/claude_mem_feed        # Show current status
/claude_mem_feed on     # Request enable
/claude_mem_feed off    # Request disable
```

### /claude_mem_status

Check worker health and session status.

```
/claude_mem_status
```

Returns worker status, port, active session count, and observation feed connection state.

## Architecture

The plugin uses HTTP calls to the already-running claude-mem worker service rather than spawning subprocesses. This means:

- No `bun` dependency required on the gateway
- No process spawn overhead per event
- Uses the same worker API that Claude Code hooks use
- All operations are non-blocking (fire-and-forget where possible)

### Session Tracking

Each OpenClaw agent session gets a unique `contentSessionId` (format: `openclaw-<sessionKey>-<timestamp>`) that maps to a claude-mem session in the worker. The plugin tracks:

- `sessionIds` — Maps OpenClaw session keys to content session IDs
- `contextCache` — TTL cache (60s) for context injection responses, keyed by project

Both are cleared on `gateway_start`.

## Requirements

- Claude-mem worker service running on `localhost:37777` (or configured port)
- OpenClaw gateway with plugin support
- Network access between gateway and worker (localhost only)