← Back to Skills Marketplace
89
Downloads
0
Stars
0
Active Installs
1
Versions
Install in OpenClaw
/install agent-gateway
Description
Smart multi-model routing — use Claude, GPT, Gemini, or local Ollama models with automatic cost optimization, fallback chains, and usage tracking.
README (SKILL.md)
Agent Gateway
You are the Claws-Shield Agent Gateway — a smart multi-model routing proxy that runs locally.
What You Do
- Multi-Provider Support — Route requests to Anthropic Claude, OpenAI GPT, Google Gemini, or local Ollama models
- Smart Routing — 3 strategies: cheapest-viable, best-quality, balanced
- Fallback Chains — If provider A fails, automatically try provider B
- Cost Tracking — Per-request cost calculation, daily/weekly/monthly aggregation
- Health Monitoring — Circuit breaker pattern, automatic provider health checking
- OpenAI-Compatible API — Drop-in replacement with
/v1/chat/completionsendpoint
How to Use
Start the gateway server:
npx @claws-shield/cli gateway --port 8787
Or programmatically:
node scripts/start-gateway.mjs
Then send requests to http://localhost:8787/v1/chat/completions using any OpenAI-compatible client.
Configuration
Set provider API keys via environment variables:
ANTHROPIC_API_KEY— For Claude modelsOPENAI_API_KEY— For GPT modelsGEMINI_API_KEY— For Gemini models- Ollama requires no key (connects to localhost:11434)
Routing Strategies
| Strategy | Description |
|---|---|
cheapest |
Pick the lowest-cost model that meets requirements |
best-quality |
Pick the most capable model available |
balanced |
Best value: quality per dollar |
Privacy
All usage data stays local. The gateway never phones home. Your API keys are never transmitted to third parties.
Usage Guidance
This skill is inconsistent: it advertises a local gateway but doesn't ship the gateway code and tells you to run 'npx @claws-shield/cli', which will fetch and execute code from npm. Before installing or running anything: 1) Ask the publisher for the upstream repository/homepage and a reproducible install artifact (repo URL, signed release, or package checksum). 2) Require the skill metadata to list the environment variables it needs (the SKILL.md references ANTHROPIC_API_KEY, OPENAI_API_KEY, GEMINI_API_KEY). 3) Audit the npm package contents (or prefer an alternative from a known source) and don't run npx on an unknown package on a production machine — run it in an isolated sandbox instead. 4) If you must provide API keys, apply least privilege and monitor usage (rotate keys after testing). 5) If the vendor claims 'never phones home', ask for source code or network logs to verify; absence of local gateway code in the skill bundle makes that claim unverifiable.
Capability Analysis
Type: OpenClaw Skill
Name: agent-gateway
Version: 1.0.0
The skill bundle is a documentation-heavy wrapper for a multi-model LLM gateway. The provided script (scripts/start-gateway.mjs) is a non-functional placeholder that directs users to a CLI tool, and the SKILL.md instructions are consistent with the stated purpose of routing and cost optimization. No malicious code, data exfiltration logic, or suspicious obfuscation was found within the provided files.
Capability Assessment
Purpose & Capability
The skill claims a full-featured multi-model gateway (routing, cost optimization, health checks, OpenAI-compatible API), but the repository only contains a short wrapper script that prints messages and a SKILL.md that instructs the user to run an external npm package (npx @claws-shield/cli). There is no code implementing the gateway included in the skill bundle and no homepage or repository URL to verify the upstream project.
Instruction Scope
SKILL.md instructs the agent/user to run 'npx @claws-shield/cli gateway --port 8787' or to run the included node script. The included script does not start a server; it only prints messages. The instructions also ask for provider API keys (ANTHROPIC_API_KEY, OPENAI_API_KEY, GEMINI_API_KEY) and assert 'privacy: never phones home' — that claim cannot be verified from the provided files. The skill's runtime instructions therefore direct the agent to fetch and execute external code and to provide credentials that are not reflected in the declared metadata.
Install Mechanism
There is no install spec in the skill itself; instead the SKILL.md tells the user to run 'npx', which dynamically downloads and executes code from the npm registry. Running npx against an unknown package can execute arbitrary code on the host. Because the skill lacks a homepage, repo, or checksum for that package, the instruction to use npx is a medium-to-high risk action.
Credentials
The skill metadata declares no required environment variables, but SKILL.md requires ANTHROPIC_API_KEY, OPENAI_API_KEY, and GEMINI_API_KEY for its functionality. This is an explicit inconsistency: the skill requests sensitive credentials in instructions but does not list them in its declared requirements. Requesting multiple provider API keys is plausible for a gateway, but the omission from metadata and the lack of included code to justify how those keys are used is suspicious.
Persistence & Privilege
The skill is not marked 'always: true' and does not request elevated or persistent system-wide privileges. The included script does not modify system configuration. Autonomous invocation is allowed (default) but is not by itself a red flag here.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install agent-gateway - After installation, invoke the skill by name or use
/agent-gateway - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.0
Initial release of Agent Gateway — a local multi-model LLM routing proxy.
- Routes LLM requests across Anthropic Claude, OpenAI GPT, Google Gemini, and local Ollama models
- Supports automatic cost optimization, fallback chains, usage tracking, and circuit breaker health monitoring
- Provides an OpenAI-compatible /v1/chat/completions API endpoint for easy integration
- Offers configurable routing strategies: cheapest, best-quality, and balanced
- All data and API keys stay on your machine for maximum privacy
Metadata
Frequently Asked Questions
What is Agent Gateway?
Smart multi-model routing — use Claude, GPT, Gemini, or local Ollama models with automatic cost optimization, fallback chains, and usage tracking. It is an AI Agent Skill for Claude Code / OpenClaw, with 89 downloads so far.
How do I install Agent Gateway?
Run "/install agent-gateway" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Agent Gateway free?
Yes, Agent Gateway is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does Agent Gateway support?
Agent Gateway is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Agent Gateway?
It is built and maintained by Blossom (@mackding); the current version is v1.0.0.
More Skills