OpenClaw Ollama Memory

Set up local semantic memory search in OpenClaw using Ollama + nomic-embed-text. Free, private, offline-capable. Replaces cloud embedding APIs (OpenAI, Gemin...

Security Scan

The skill is internally coherent: it documents how to configure OpenClaw to use a local Ollama embedding model, requires only Ollama and curl, and contains no hidden endpoints or secret-exfiltration steps.

Purpose & Capability

The name/description match the instructions: it configures OpenClaw to use Ollama + nomic-embed-text. One small inconsistency: the runtime instructions call 'openclaw gateway restart' but the skill's declared required binaries list does not include the openclaw CLI. That omission is minor (the skill necessarily assumes OpenClaw is already installed), but the registry metadata should list openclaw if the skill expects it to be present.

Instruction Scope

Instructions stay within the stated purpose (pull a local model, point OpenClaw at a local Ollama API, and restart the gateway). They reference editing ~/.openclaw/openclaw.json and optionally adding 'extraPaths' (absolute or relative) which OpenClaw will scan recursively for .md files — this is expected for a memory-indexing feature but carries the usual risk that if you configure extraPaths to system or sensitive directories, those files may be indexed. The SKILL.md does not instruct reading arbitrary system files or contacting external endpoints beyond ollama.ai (download page) and the local Ollama API at 127.0.0.1.

Install Mechanism

This is instruction-only (no install spec). It points users to download Ollama from the official site and to run 'ollama pull'. No archives or third-party download URLs are embedded in the skill itself. This keeps the install risk low, though users should verify they download Ollama from the official domain.

Credentials

The skill requests no environment variables or credentials and only requires local binaries (ollama, curl). That is proportionate for a local embedding provider. No unrelated cloud API keys or secret variables are requested.

Persistence & Privilege

The skill does not request always: true or any elevated platform privileges. It is user-invocable and may be invoked autonomously per platform defaults, which is appropriate for an integration helper. It does not attempt to modify other skills' configs; it only instructs the user to update OpenClaw's own config file.

Assessment

This skill appears to do what it says: configure OpenClaw to use a local Ollama embedding model. Before installing: 1) ensure you have OpenClaw installed (the SKILL.md expects the 'openclaw' CLI) and Ollama downloaded from the official site; 2) keep Ollama bound to localhost (127.0.0.1) as recommended to avoid exposing the local embedding API to the network; 3) be cautious when adding extraPaths — pointing OpenClaw at arbitrary absolute directories can cause sensitive files to be indexed; and 4) verify disk/CPU requirements for pulled models. If you want the registry metadata to be strictly accurate, ask the publisher to add 'openclaw' to the required binaries list.

Local semantic memory search for OpenClaw using Ollama — free, private, no API key required.

Quick Start

1. Install & Run Ollama

Download from https://ollama.ai, then pull the embedding model:

ollama pull nomic-embed-text

Verify Ollama is running:

curl http://127.0.0.1:11434/api/tags

2. Configure OpenClaw

Edit ~/.openclaw/openclaw.json and add under agents.defaults.memorySearch:

{
  "agents": {
    "defaults": {
      "memorySearch": {
        "provider": "ollama",
        "model": "nomic-embed-text:latest",
        "remote": {
          "baseUrl": "http://127.0.0.1:11434"
        }
      }
    }
  }
}

3. Restart OpenClaw

openclaw gateway restart

Key Points

• Ollama must be explicit — use provider: "ollama", not "openai". OpenClaw doesn't auto-detect Ollama.
• No API key needed — runs locally, fully offline.
• nomic-embed-text — 274MB model, fast, accurate, free.
• baseUrl format — http://127.0.0.1:11434 (no /v1 suffix, no trailing slash).

Next Steps

• See references/config-reference.md for advanced tuning (hybrid search, MMR, temporal decay, extra paths).
• See references/troubleshooting.md for common errors and fixes.