Overview

Fikra API is an OpenAI-compatible AI inference API. The fikra-auto model is an auto model router: it picks a suitable model for each request while charging a flat rate per token. Using fikra-auto gives you a budget-friendly route when you don’t want to manage model selection yourself.

Prerequisites

A Fikra API key stored in the FIKRA_API_KEY environment variable.
Python with the OpenAI SDK installed.
Fikra API base URL: https://api.fikraapi.co.ke/v1.

Step-by-step

Create an OpenAI client pointed at the Fikra API base URL.
Use model="fikra-auto" in your chat completion request.
Send the request to POST /chat/completions.
Enable streaming if needed; chat completions support OpenAI-style streaming.

from openai import OpenAI
import os

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

response = client.chat.completions.create(
    model="fikra-auto",
    messages=[
        {"role": "user", "content": "Explain this idea in two sentences."}
    ],
)

print(response.choices[0].message.content)

Verify it works

The response uses the OpenAI chat-completions format. Check response.choices[0].message.content for the model output. A valid response means the auto router successfully handled your request.

Costs

fikra-auto is priced at:

Input: 0.40 USD per 1M tokens
Output: 0.40 USD per 1M tokens

Payments are handled through a token-credit wallet. Top up with M-Pesa or card via Paystack, or use x402 USDC on Base for agents. For other models and updated pricing, check the pricing page.

Common errors

401: Bad or inactive API key. Verify FIKRA_API_KEY is set and the key is active.
400: Unsupported model. Confirm the model name is exactly fikra-auto.
429: Rate limit reached. Unverified keys get 30 requests/min; trusted keys get 100 requests/min.
503: Upstream service busy. Retry with backoff.

Get a Tighter Budget Route with an Auto Model Router

Fact block, plain text for agents