OpenClaw + WisGate configuration hub

OpenClaw Zone

Connect WhatsApp, Telegram, Discord, Slack, and other channels to a self-hosted AI agent gateway, then use WisGate to manage models, cost, and fallback strategy.

Get WisGate API Key Read OpenClaw Docs

Multi-channel

Chat, tool, and automation entrypoints

/v1

OpenAI-compatible endpoint

2 setups

OpenClaw and Hermes configs

Start with the role

OpenClaw is a self-hosted AI assistant gateway

It is not a single scraping model. It connects message channels, tools, scheduled jobs, and LLMs into a local-first agent framework.

Self-hosted first

Run the core service in your own environment so data, keys, logs, and permission boundaries stay under your control.

Multi-channel access

Unify WhatsApp, Telegram, Discord, Slack, iMessage, and similar entrypoints into the same agent workflow.

Model neutral

Connect OpenAI-compatible endpoints, local models, or frontier models, and route each task to the right capability and price.

Automation-ready

Use cron, webhooks, tool calls, and multi-step jobs for always-on assistants and team automation.

Why add WisGate

One model gateway for OpenClaw and Hermes

WisGate provides an OpenAI-compatible API. Point your agent at one endpoint, then choose Claude, GPT, DeepSeek, Gemini, or other available models by task.

Recommended endpoint

https://api.wisgate.cn/v1

Reduce key sprawl

Consolidate multi-provider key management into one WisGate API key that is easier to share, rotate, and audit.

Works with existing agents

Use the OpenAI-compatible `/v1` API with tools such as OpenClaw and Hermes that support custom endpoints.

Switch models by task

Daily chat, code repair, long-context research, and low-cost batch work can use different models instead of one fixed subscription.

Simpler fallback

Configure backup models when a frontier model is unavailable or too expensive, reducing downtime for always-on agents.

Use case map

The jobs that belong in a Claw zone

The page should help users decide what to run with OpenClaw and when Hermes is worth adding.

OpenClaw

Personal always-on assistant

Receive messages across chat apps, organize reminders, answer contextual questions, and connect daily tools.

OpenClaw

Team collaboration entrypoint

Bring project Q&A, task handoff, knowledge retrieval, and engineering automation into Slack or Discord.

OpenClaw + Hermes

Research and summaries

Track news, earnings, product updates, or technical articles on a schedule, then push summaries to the team.

OpenClaw

DevOps and self-healing

Receive alerts, run diagnostics, trigger webhooks or n8n flows, and handle repetitive operations work.

OpenClaw

Social listening

Monitor X, Reddit, YouTube, or community keywords, then summarize trends and risk signals.

OpenClaw + Hermes

Creative pipelines

Split research, script drafting, image generation, and publishing checks into multiple agent steps.

OpenClaw quick setup

From onboarding to safe launch

Give users an executable path: initialize, configure the WisGate endpoint, then verify permissions and connectivity.

Initialize OpenClaw

Run `openclaw onboard` to complete first-time setup, generate local configuration, and choose channels.

Edit the OpenClaw config

Keep primary settings in `~/.openclaw/openclaw.json`, and inject secrets through environment variables instead of committing them.

Connect the WisGate provider

Use `https://api.wisgate.cn/v1` as the OpenAI-compatible base URL, then set a default model and fallback models.

Set allowlists and webhook tokens

Limit the agent to trusted users or groups, and use a dedicated token for webhook access.

Run checks before launch

Use `openclaw doctor` to verify configuration, connectivity, and permissions before testing in a small channel.

Example configuration

~/.openclaw/openclaw.json

{
  "llm": {
    "provider": "openai-compatible",
    "baseUrl": "https://api.wisgate.cn/v1",
    "apiKey": "$WISGATE_API_KEY",
    "model": "claude-opus-4-7",
    "fallbackModels": ["gpt-5.5", "deepseek-v4-pro"]
  },
  "security": {
    "allowedUsers": ["you@example.com"],
    "webhookToken": "$OPENCLAW_WEBHOOK_TOKEN"
  }
}

Common check commands

openclaw onboard
openclaw doctor
hermes model

Hermes setup clarified

A clear path for Hermes + WisGate

Keep Hermes as the supporting setup path inside the OpenClaw zone: connect WisGate through the official wizard, then make the YAML, secrets, and checks easy to reuse.

Run the Hermes model wizard

Run `hermes model`, choose `Add Custom Endpoint` or `Custom endpoint`, and enter the WisGate base URL, API key, and model name.

Persist it in config.yaml

After the wizard works, turn the setup into a team template in `~/.hermes/config.yaml` with the primary provider set to `custom`.

Enter the WisGate endpoint

Use `https://api.wisgate.cn/v1` as the default endpoint.

Reference secrets as variables

Use `api_key: $` in YAML, then inject the key through `.env` or your deployment environment.

Validate in a new session

Run `hermes config check`, then test in a new session. If gateway mode is running, restart the gateway after changing config.

Hermes manual config

~/.hermes/config.yaml

model:
  default: claude-opus-4-7
  provider: custom
  base_url: https://api.wisgate.cn/v1
  api_key: ${WISGATE_API_KEY}

auxiliary:
  compression:
    provider: auto
    model: ''
  vision:
    provider: auto
    model: ''

terminal:
  backend: docker
  cwd: "."
  timeout: 180

Prefer `hermes model` for Custom endpoint setup. When editing manually, use `~/.hermes/config.yaml`. Inject API keys through `~/.hermes/.env` or your deployment environment instead of hard-coding them into shareable files.

Validation and troubleshooting

hermes model
hermes config check
hermes gateway restart

How to choose

OpenClaw vs Hermes

They are not simple substitutes. OpenClaw is the entry and orchestration layer; Hermes is the long-running executor that can accumulate skills.

Dimension	OpenClaw	Hermes
Best for	Multi-channel personal assistants, team entrypoints, and chat-driven automation.	Repeated tasks, self-improvement, long-term skill accumulation, and batch execution.
Core strength	Connects channels, tools, webhooks, cron jobs, and model routing.	Turns experience into reusable skills for workflows that improve over time.
Deployment mindset	Stabilize the gateway and channels first, then add automation gradually.	Define the task environment and model config first, then observe execution quality.
Memory and learning	Focused on message context, tool context, and session orchestration.	Focused on long-running task experience, skill generation, and self-optimization.
WisGate role	A unified LLM endpoint shared by multiple channels and model strategies.	A Custom endpoint that gives Hermes a stable model gateway.

Launch safety checklist

Keep always-on agents controlled, auditable, and recoverable

Keep secrets out of git

Inject API keys, webhook tokens, and chat platform tokens through environment variables or secret managers.

Restrict users and channels

Start with trusted accounts or internal channels, then expand based on logs and review.

Authenticate webhooks separately

Use a dedicated webhook token and log source, time, and execution result.

Give tools minimum permissions

File, browser, terminal, and third-party API access should be scoped to the task instead of open by default.

Keep audit logs

Record the trigger source, input summary, tool calls, model response, and execution result so mistakes can be traced and rolled back.

Check connectivity before launch

Verify configuration, model access, networking, and permissions; if the tool offers `doctor` or `config check`, use it as part of pre-launch validation.

Deep Dive

Explore the configuration details, Hermes self-improvement model, and practical agent workflow decisions in more depth.

OpenClaw Complete Configuration Guide

Use 36 scenarios to understand model routing, cost optimization, and real automation patterns.

Hermes Agent Complete Guide

Learn how Hermes handles self-improvement, skill accumulation, and the tasks it fits best.

Hermes vs OpenClaw

Use the selection table to decide which agent fits your workflow.