OpenClaw Visual Config.yaml Generator
Configure agents, model parameters, tools, sandbox permissions, and vector memory with real-time YAML generation.
version: '1.0'
agent:
name: OpenClaw-Assistant
model: gpt-4o
temperature: 0.7
max_tokens: 4096
system_prompt: You are an intelligent, helpful AI assistant built on OpenClaw. Execute tasks methodically and verify results.
tools:
web_search:
enabled: true
max_results: 5
timeout_ms: 8000
code_interpreter:
enabled: true
sandbox: docker
filesystem:
enabled: true
root_dir: ./workspace
read_only: false
bash:
enabled: false
timeout_ms: 10000
browser:
enabled: false
headless: true
memory:
backend: sqlite
connection_uri: ./data/memory.sqlite
embedding_model: text-embedding-3-small
retention_days: 30
runtime:
log_level: info
max_concurrency: 2
rate_limit_rpm: 60
port: 8080
Ready to deploy this agent on a cloud VPS?
Run OpenClaw 24/7 with dedicated CPU, NVMe SSDs, and instant Docker deployment on Vultr worldwide cloud edge nodes.
Deploy on Vultr Cloud VPSLearn how to install, troubleshoot, and scale OpenClaw in production:
Frequently Asked Questions about OpenClaw Configuration
Where should I place the generated `config.yaml` file?
Place `config.yaml` in the root directory of your OpenClaw workspace or pass it via the CLI flag:openclaw --config ./config.yaml. OpenClaw will automatically validate the YAML schema on startup.
How do I supply API keys securely without putting them in `config.yaml`?
Never commit plaintext API keys into version control. OpenClaw automatically reads environment variables (such asOPENAI_API_KEY,ANTHROPIC_API_KEY, orDATABASE_URL) from your `.env` file or server environment.
Can I run OpenClaw with local offline LLMs using Ollama?
Yes. Select the Local Privacy / Ollama preset above. Ensure your local Ollama server is running (ollama serve) and set the model identifier to your local model name (e.g., llama3.3:70b).
What is the difference between SQLite and PostgreSQL memory backends?
SQLite is embedded and requires zero external infrastructure, making it ideal for single-agent and local development. PostgreSQL with pgvector enables high-concurrency vector search, multi-agent shared memory, and scalable production deployments.
Universal Model Support
Seamlessly configure OpenAI GPT-4o, Anthropic Claude 3.5, DeepSeek R1/V3, or self-hosted Ollama and vLLM inference endpoints with unified parameters.
Sandboxed Tooling
Control filesystem write boundaries, restrict dangerous terminal commands, and define isolated Docker/gVisor microVM sandbox runtimes.
1-Click Production Ready
Generate 100% syntactically valid YAML verified for OpenClaw runtime standards. Instant clipboard copy and file download.