For OpenClaw & Hermes Agents

Supercharged Agents.Up to 75% Savings.One Command Line.

Drop the Yielding Bear skill into any OpenClaw or Hermes session. Every LLM call — summarization, code review, reasoning, support reply — routes through the most efficient model automatically. Up to 75% savings vs retail.

Get Free API Key →How it works

Set up in 60 seconds

No SDK rewrite. No model selection logic. Just one env var.

Install the skill

Run install yieldingbear in any OpenClaw or Hermes session. Drops the routing hooks in automatically.

install yieldingbear

Set your API key

Add your free Yielding Bear key as an env var. Routing kicks in on the next call.

YIELDINGBEAR_API_KEY=***

Save on every call

Simple lookups → Llama 8B / Gemini Flash / GPT-4o-mini. Hard reasoning → Claude Sonnet / GPT-4o. No prompt changes, no SDK work, no per-call decisions.

Two real agents. Real daily bills.

We measured 30 days of production traffic on two flagship agents. Here's what YieldingBear routing did to each one.

OpenClaw agents

Discord ops, mobile builds, customer support

OpenClaw runs 200+ LLM calls per session — Discord replies, build logs, support answers, code reviews. Every one of them gets routed through the unified API automatically. No per-call model selection, no prompt changes.

Daily bill before:$14.20

After Yielding Bear:$2.10

85% saved

~240 calls/day

Setup

// In your OpenClaw session

install yieldingbear

export YIELDINGBEAR_API_KEY="***"

// Done. Every LLM call from this session now routes through Grizzly 1.0G.

Hermes agents

Deep research, multi-step investigations, code gen

Hermes agents run long-form research and engineering work — hundreds of calls per task, often across 30+ minutes. Routing those through the unified API means simple lookups hit Llama 8B and only the hard reasoning lands on Claude/GPT-4o.

Daily bill before:$22.40

After Yielding Bear:$3.80

83% saved

~380 calls/day

Setup

yieldingbear.register(api_key="***")

// Or in your .env:

YIELDINGBEAR_API_KEY="***"

// All Hermes LLM calls auto-route. Override per-call with model= if needed.

What changes in your code

One line. That's it.

Beforealways-Claude, always full price

from anthropic import Anthropic

client = Anthropic(
    api_key=os.environ["ANTHROPIC_API_KEY"],
)

resp = client.messages.create(
    model="claude-3-5-sonnet-latest",
    max_tokens=1024,
    messages=[{"role": "user", "content": prompt}],
)

Every call is $3/$15 per 1M tokens. Summarization, classification, formatting — all priced the same.

AfterYielding Bear routes to the most efficient model

from openai import OpenAI

client = OpenAI(
    base_url="https://yieldingbear.com/v1",
    api_key=os.environ["YIELDINGBEAR_API_KEY"],
)

resp = client.chat.completions.create(
    model="grizzly-1.0g",  # ← that's it
    max_tokens=1024,
    messages=[{"role": "user", "content": prompt}],
)

Simple lookups land on Llama 8B. Hard reasoning escalates to Claude/GPT-4o. Net cost: ~65% lower on typical traffic.

Start saving on your agent's LLM bill today

Free API key, no credit card. 50K tokens/mo on the house. Your first routed call lands in <60 seconds.

Get Free API Key →Read the Docs