We built Fikra API because African developers were systematically locked out of the global artificial intelligence boom. Between international credit card walls, steep $20 minimum top-ups, and the latency of routing requests to servers in Virginia or Dublin, building AI applications locally was both economically and technically unviable. Today, we are changing that paradigm. We are thrilled to announce the official release of the Fikra API Developer Documentation, giving you the keys to drop world-class AI models directly into your stack in minutes.
The Engineering Philosophy: A True Drop-In Replacement
When we set out to build the Fikra infrastructure, our primary goal was to eliminate friction. We knew that developers in Nairobi, Lagos, and Cairo had already experimented with global APIs. Forcing them to learn a new, proprietary payload structure would be a massive barrier to entry. That is why the Fikra API was engineered from the ground up to be a 1:1 drop-in replacement for the OpenAI standard.
If you know how to build with the most popular global AI APIs, you already know how to build with Fikra. The parameter names, the JSON response structures, and the streaming protocols are identical. The only things that change are the base URL and the speed at which your local requests resolve.
import openai
# Migration takes less than 30 seconds
client = openai.OpenAI(
base_url="https://api.fikraapi.co.ke/v1",
api_key="your_fikra_key"
)
response = client.chat.completions.create(
model="fikra-fast-8b",
messages=[{"role": "user", "content": "Explain the mechanics of mobile money integrations in Python."}]
)
print(response.choices[0].message.content)Under the Hood: M-Pesa Native Billing Infrastructure
The most significant bottleneck for African SaaS is payments. To solve this, the Fikra API documentation introduces our proprietary M-Pesa native billing architecture. We have bypassed the Visa/Mastercard networks entirely to create a micro-transaction ledger that operates in Kenyan Shillings (KES).
Developers can fund their inference accounts directly via an M-Pesa STK push. The conversion rate is currently fixed at an astonishingly accessible tier: approximately 2,000,000 tokens per $1 equivalent, with a minimum top-up of just KES 130. This means a university student can build, test, and deploy a generative AI application over the weekend using the pocket change on their mobile phone.
Model Routing and Fleet Capabilities
The documentation outlines our initial fleet of models, tailored for different enterprise requirements. Our routing layer is designed for high-availability across our East African edge nodes.
fikra-fast-8b: Our flagship high-throughput model. Utilizing 1.58-bit ternary weight quantization, this model achieves blistering inference speeds with minimal VRAM footprint. It is ideal for real-time chatbots, rapid text classification, and user-facing conversational interfaces where latency is the defining metric.fikra-pro-120b: Designed for complex reasoning, long-context summarization, and zero-shot code generation. While slightly slower, its parameter count allows it to handle nuanced logic puzzles and complex systemic workflows that smaller models struggle with.fikra-embedding-v1: Recently launched for high-dimensional vectorization, perfectly suited for developers building Retrieval-Augmented Generation (RAG) pipelines and semantic search databases.
Enterprise Integrations and Data Sovereignty
For enterprise clients, the documentation provides extensive blueprints for scale. We have included comprehensive guides on setting up asynchronous webhooks and directly connecting the Fikra API to the WhatsApp Business API—a crucial integration for conversational commerce on the continent.
Furthermore, by routing requests through our local infrastructure, enterprises finally have a solution for data sovereignty. Financial institutions, healthcare providers, and government agencies can now deploy AI solutions knowing their sensitive payload data never leaves the region or hits overseas servers.
Start Building the Future
The infrastructure is deployed. The network is stable. The documentation is written. Now, it is up to the builders. We are incredibly excited to see what the next generation of African engineers will create when the barriers to global-tier artificial intelligence are finally removed.
The official documentation is live, open, and actively maintained. Grab your API keys and start building today at docs.fikraapi.co.ke.
