Get an AI key and add it to MeAPI

MeAPI’s AI features — the AI chatbot, knowledge-base answers, suggested replies, summaries and translation — run on your own AI provider account. You pay the provider directly at their price; MeAPI adds nothing on top.

Which provider?

Provider	Good for	Knowledge-base search by meaning	Free tier
Google Gemini	Low cost, strong in Indian languages	✅	Yes (rate-limited)
OpenAI	All-round quality	✅	No — prepaid credit
Anthropic Claude	Careful, well-written answers	— (pair with another for embeddings)	No — prepaid credit
Groq	Very fast, cheap open models	—	Yes (rate-limited)
Mistral	Low cost, European	✅	Limited
DeepSeek	Very low cost	—	No
xAI Grok	Grok models	—	No
OpenRouter	One key for hundreds of models	—	Some free models
Azure OpenAI	Companies already on Azure	✅	Azure billing

Not sure? Start with Google Gemini — it has a free tier, handles Hindi and Gujarati well, and also powers knowledge-base search. If you use Claude, Groq or DeepSeek for answers, add a Gemini, OpenAI or Mistral key as well so your knowledge bases can be searched by meaning (otherwise they use keyword search).

Create a key

Gemini

1. Open aistudio.google.com/apikey and sign in with a Google account.
2. Click Create API key, pick or create a Google Cloud project.
3. Copy the key (starts with AIza).
4. Optional: to go beyond the free tier’s limits, enable billing on the project in Google Cloud.

OpenAI

1. Open platform.openai.com and sign in.
2. Add credit under Settings → Billing (the API is prepaid; a ChatGPT subscription does not include API use).
3. Go to API keys → Create new secret key, give it a name such as “MeAPI”.
4. Copy the key (starts with sk-). Set a monthly budget under Limits if you like.

Claude

1. Open console.anthropic.com and sign up.
2. Buy credits under Settings → Billing (a Claude.ai subscription does not include API use).
3. Go to Settings → API keys → Create key, name it “MeAPI”.
4. Copy the key (starts with sk-ant-).

Groq

1. Open console.groq.com/keys and sign in.
2. Click Create API Key and copy it (starts with gsk_).
3. The free tier is rate-limited; upgrade in the console for production volume.

Others

• Mistral: console.mistral.ai → choose a plan → API keys → Create new key.
• DeepSeek: platform.deepseek.com → top up → API keys → Create new API key.
• xAI Grok: console.x.ai → add credits → API keys → Create API key.
• OpenRouter: openrouter.ai/keys → Create key; add credits for paid models. Model names look like openai/gpt-4.1-mini.
• Azure OpenAI: in the Azure portal create an Azure OpenAI resource and deploy a chat model (and an embedding model). You need the endpoint (https://<resource>.openai.azure.com), a key from Keys and Endpoint, and the deployment names.
• Custom: any OpenAI-compatible server (for example a self-hosted model gateway): its base URL ending in /v1 and a key. It must be reachable over public https.

Add the key in MeAPI

1. In the panel open AI & Knowledge → AI providers (account owner only).
2. Click the provider, paste the key (and, for Azure / Custom, the endpoint and deployment names).
3. Click Save & test. MeAPI sends a tiny test request and shows whether the key works and how fast it answered. For providers with embeddings it tests those too.
4. Pick a default model — Load models fetches the list from the provider. Cheaper “mini / flash / haiku” models are plenty for customer replies.
5. If you have more than one key, choose which one is the default and which one does embeddings under Which key does what.

Then set limits under AI & Knowledge → Usage & limits: a monthly spend cap and a cap on AI replies per conversation per day. When a cap is hit, AI stops and chats go to your team, and you get an alert.

Tip

Keep your key safe

• MeAPI encrypts the key and never shows it again — not even to you, and never to your agents.
• Create a key just for MeAPI, so you can revoke it without affecting anything else.
• Also set a spending limit at the provider; it is the last line of defence.
• If the provider rejects the key or your credit runs out, AI replies pause, chats go to your team and the owner gets an alert. Fix the key and AI resumes.

What does it cost?

AI providers charge per token (roughly ¾ of a word). A typical WhatsApp answer from a knowledge base uses about 1,500–3,000 tokens including the context, which costs roughly ₹0.03–₹0.15 on small models like Gemini Flash, GPT-4.1 mini or Claude Haiku, and more on large models. AI & Knowledge → Usage & limits shows tokens and estimated cost per day, provider, bot and feature, so you always know.