Skip to main content
This page shows the fastest path: Proxy mode with the OpenAI SDK. Other modes are linked at the bottom.

​
1. Get your PromptWall API key

Sign in at www.prompt-wall.com, finish onboarding, and copy the pw_… key shown on the final step. Save it as PROMPTWALL_API_KEY in your environment.

​
2. Point your OpenAI client at PromptWall

You change two things:
  1. OPENAI_BASE_URL → https://api.prompt-wall.com/v1
  2. Add a X-PromptWall-Api-Key default header.
Your OpenAI key stays where it always was. 
from openai import OpenAI

client = OpenAI(
    base_url="https://api.prompt-wall.com/v1",
    api_key=os.environ["OPENAI_API_KEY"],
    default_headers={
        "X-PromptWall-Api-Key": os.environ["PROMPTWALL_API_KEY"],
        "X-Use-Case":           "customer_support",
    },
)

resp = client.chat.completions.create(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

​
3. Watch it on the dashboard

Open www.prompt-wall.com/observability. Within seconds your request shows up:
  • KPI cards update (requests, cost, p95 latency, error rate)
  • Filter by use_case = customer_support to slice by the dimension you tagged
  • Click any row in Recent traces to see the full waterfall

​
4. Turn on runtime control (optional)

Default mode is observe — capture only, no mutations. To turn on control, send X-PromptWall-Mode per request or set policy_json.proxy_control_mode per tenant.
Header valuePre-LLMPost-LLM
observe——
guardguards + injection scan—
verify—PII redaction
fullbothboth

​
Other integration modes