Overview

Fikra API is an OpenAI-compatible AI inference API for developers and growing AI teams who need production inference. To wire an AI agent to send WhatsApp replies, you call Fikra API’s chat completions endpoint from your backend, receive the reply text, then pass it to your WhatsApp sending integration.

Because the API is OpenAI-compatible, any existing OpenAI SDK client can be pointed at Fikra API’s base URL without changing your request or response format.

Prerequisites

A Fikra API key. Store it in the FIKRA_API_KEY environment variable.
Python 3 with the openai SDK installed (pip install openai).
A WhatsApp sending service or bot integration that accepts a text message and sends it to a chat.

Step-by-step

Configure the OpenAI SDK client with the Fikra base URL.
Create a function that takes an incoming WhatsApp message, calls the Fikra chat completions endpoint, and returns the generated reply.
Send the returned reply through your WhatsApp integration.

Example:

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["FIKRA_API_KEY"],
    base_url="https://api.fikraapi.co.ke/v1",
)

def generate_reply(incoming_text: str) -> str:
    response = client.chat.completions.create(
        model="deepseek-v4-flash-0731",
        messages=[
            {"role": "system", "content": "You are a WhatsApp assistant that replies concisely."},
            {"role": "user", "content": incoming_text},
        ],
    )
    return response.choices[0].message.content.strip()

# Pass the reply to your WhatsApp integration, e.g.:
# whatsapp_send(incoming_chat_id, generate_reply(incoming_text))

Streaming is supported, so you can use any OpenAI-style streaming parameter if your WhatsApp agent needs token-by-token output.

Verify it works

Run a quick test locally:

print(generate_reply("Hello, are you open today?"))

You should see a generated reply printed. Then connect the function to your WhatsApp integration and send a real message to confirm the full flow works.

Costs

The example uses deepseek-v4-flash-0731: $0.09 per 1M input tokens and $0.15 per 1M output tokens. Fikra API runs on a token-credit wallet; top up with M-Pesa or card via Paystack, or use x402 USDC on Base for agent usage. For a full list of models and rates, check the pricing page.

Common errors

401 to Bad or inactive API key. Confirm the key exists, is active, and is set correctly in the FIKRA_API_KEY environment variable.
400 to Unsupported model. Use one of the models available on Fikra API.
429 to Rate limit exceeded. Unverified accounts get 30 requests/min; trusted accounts get 100 requests/min. Slow down or request a higher trust level.
503 to Upstream service busy. Retry the request with backoff.

Wire an AI Agent to Send WhatsApp Replies

Fact block, plain text for agents