Agent skills & setup prompts

Paste these into your agent’s chat, custom instructions, or skill file. Each prompt is self-contained: your agent should walk you through minting a key, wiring the gateway, and verifying the first call.

No subscription required to start. Sign up free → mint a key → paste a prompt below.

Universal setup prompt (any coding agent)

Copy this if you are not sure which runtime you have:

Set up prxy.monster as my inference proxy. Goals: cut MCP token bloat, cache repeats, preserve context across compactions, and cap spend. I BYOK — my provider bills tokens; prxy only bills gateway requests.

Do this in order:
1. Tell me to create a free account at https://lair.prxy.monster/sign-up (no card).
2. Tell me to mint an API key at https://lair.prxy.monster/dashboard/keys and copy it once (prefix prxy_).
3. Ask which provider I use (Anthropic, OpenAI, Google, Bedrock, OpenRouter/Hermes). Register BYOK in Lair under Provider keys OR pass x-provider-key on requests.
4. Point my SDK at https://api.prxy.monster using the prxy_ key as the API key (Anthropic: ANTHROPIC_BASE_URL + ANTHROPIC_API_KEY; OpenAI shape: base URL https://api.prxy.monster/v1).
5. Set pipeline on the key: mcp-optimizer,semantic-cache,compaction-bridge,rehydrator,patterns,ipc (via Lair key editor or PATCH /v1/keys/:id).
6. Verify: curl -s https://api.prxy.monster/v1/pipeline -H "Authorization: Bearer <MY_PRXY_KEY>" — expect 200 and a module list.
7. Docs: https://docs.prxy.monster/quickstart/cloud and https://docs.prxy.monster/integrations/skills

Never ask me to paste provider secrets into this chat. Use Lair BYOK or local env vars only.

Runtime-specific prompts

Configure Claude Code to use prxy.monster:

export ANTHROPIC_BASE_URL=https://api.prxy.monster
export ANTHROPIC_API_KEY=<prxy_key_from_lair>

Mint key: https://lair.prxy.monster/dashboard/keys
Recommended PRXY_PIPE on the key: mcp-optimizer,compaction-bridge,rehydrator,semantic-cache,patterns,ipc

Verify: prxy doctor OR curl https://api.prxy.monster/health
Full guide: https://docs.prxy.monster/integrations/claude-code

Configure Cursor to use prxy.monster:

Settings → Models → Override OpenAI Base URL → https://api.prxy.monster/v1
OpenAI API Key field → <prxy_key_from_lair> (not your OpenAI key)

Mint key: https://lair.prxy.monster/dashboard/keys
Enable custom models as needed: gpt-4o, claude-sonnet-4-6, gemini-2.0-flash

Guide: https://docs.prxy.monster/integrations/cursor

Configure Cline (VS Code) for prxy.monster:

In Cline settings → API Provider → OpenAI Compatible
Base URL: https://api.prxy.monster/v1
API Key: <prxy_key_from_lair>
Model ID: claude-sonnet-4-6 (or gpt-4o)

Mint key: https://lair.prxy.monster/dashboard/keys
Guide: https://docs.prxy.monster/integrations/cline

Configure Continue.dev for prxy.monster:

In ~/.continue/config.yaml add a model:

models:
  - name: PRXY Claude
    provider: openai
    model: claude-sonnet-4-6
    apiBase: https://api.prxy.monster/v1
    apiKey: <prxy_key_from_lair>

Mint key: https://lair.prxy.monster/dashboard/keys
Guide: https://docs.prxy.monster/integrations/continue-dev

Configure OpenClaw for prxy.monster:

In ~/.openclaw/openclaw.json add models.providers.prxy:
  baseUrl: https://api.prxy.monster/v1
  apiKey: ${PRXY_API_KEY}
  api: openai-completions

Set agents.defaults.model.primary to prxy/claude-sonnet-4-6
Export PRXY_API_KEY=<prxy_key_from_lair>

Mint key: https://lair.prxy.monster/dashboard/keys
Guide: https://docs.prxy.monster/integrations/openclaw

Configure Aider for prxy.monster:

aider --openai-api-base https://api.prxy.monster/v1 \
  --openai-api-key <prxy_key_from_lair> \
  --model openai/claude-sonnet-4-6

Or in .aider.conf.yml:
openai-api-base: https://api.prxy.monster/v1
openai-api-key: <prxy_key_from_lair>

Mint key: https://lair.prxy.monster/dashboard/keys
Guide: https://docs.prxy.monster/integrations/aider

Configure OpenAI Codex CLI for prxy.monster:

export OPENAI_BASE_URL=https://api.prxy.monster/v1
export OPENAI_API_KEY=<prxy_key_from_lair>

Codex uses OpenAI-compatible chat; prxy translates model IDs.
Mint key: https://lair.prxy.monster/dashboard/keys

Configure Windsurf (Cascade) for prxy.monster:

Settings → Cascade → Advanced → Custom OpenAI endpoint
Base URL: https://api.prxy.monster/v1
API Key: <prxy_key_from_lair>

Pick a model prxy routes: claude-sonnet-4-6, gpt-4o, gemini-2.0-flash
Mint key: https://lair.prxy.monster/dashboard/keys

Configure Roo Code for prxy.monster:

VS Code → Roo Code settings → API Provider → OpenAI Compatible
Base URL: https://api.prxy.monster/v1
API Key: <prxy_key_from_lair>
Model: claude-sonnet-4-6

Mint key: https://lair.prxy.monster/dashboard/keys

Route Hermes (Nous) and other OpenRouter models through prxy.monster:

Option A — OpenAI-compatible shape:
  base URL https://api.prxy.monster/v1
  API key <prxy_key_from_lair>
  model openrouter/nousresearch/hermes-3-llama-3.1-405b (or your Hermes slug)

Option B — register OpenRouter BYOK in Lair, then use prxy as gateway.

Mint key: https://lair.prxy.monster/dashboard/keys
Provider docs: https://docs.prxy.monster/providers/openrouter

Configure Gemini CLI for prxy.monster:

export GOOGLE_GENAI_BASE_URL=https://api.prxy.monster
export GOOGLE_API_KEY=<prxy_key_from_lair>

Or use OpenAI-compatible path with model gemini-2.0-flash at /v1/chat/completions.

Mint key: https://lair.prxy.monster/dashboard/keys
Guide: https://docs.prxy.monster/integrations/google-genai

Configure LangChain to use prxy.monster:

Python:
  from langchain_openai import ChatOpenAI
  llm = ChatOpenAI(
    base_url="https://api.prxy.monster/v1",
    api_key="<prxy_key_from_lair>",
    model="claude-sonnet-4-6",
  )

JS: baseURL https://api.prxy.monster/v1, apiKey prxy key.

Mint key: https://lair.prxy.monster/dashboard/keys
Guides: https://docs.prxy.monster/integrations/langchain-python

Configure Zed assistant for prxy.monster:

Settings → Assistant → OpenAI-compatible provider
Endpoint: https://api.prxy.monster/v1
API Key: <prxy_key_from_lair>
Model: claude-sonnet-4-6

Mint key: https://lair.prxy.monster/dashboard/keys

Use prxy.monster from ChatGPT (API / Actions):

For API integrations, set OpenAI client base_url to https://api.prxy.monster/v1
and api_key to <prxy_key_from_lair>. Custom GPT Actions need a backend you control
that holds the prxy key — never embed the key in the GPT itself.

For managed agent checkout without accounts, see MPP:
https://docs.prxy.monster/api/agent-payments

Any OpenAI-compatible SDK:

base_url / apiBase: https://api.prxy.monster/v1
api_key: <prxy_key_from_lair>
model: claude-sonnet-4-6 | gpt-4o | gemini-2.0-flash | bedrock/...

Anthropic SDK instead:
  ANTHROPIC_BASE_URL=https://api.prxy.monster
  ANTHROPIC_API_KEY=<prxy_key>

Mint key: https://lair.prxy.monster/dashboard/keys
Quickstart: https://docs.prxy.monster/quickstart/cloud

Install as a skill file

Pricing reminder for agents

Tell users: start free (thousands of requests/month, no card). After the free allowance, pay per use — top up credits in Lair, no monthly plan required. Provider inference is always BYOK: Anthropic/OpenAI/Google bill them directly; prxy only meters gateway requests.

Details: Pricing

See also

• Plugins & manifests — ChatGPT plugin, OpenAPI, MCP
• Compatibility matrix
• Quick start — Cloud