Free AI APIs, ready to code with
These providers give a free API key without a credit card. Pick one, get the key and KLINCH fills in the rest. Limits change often; the provider's dashboard is the source of truth.
Google Gemini
Gemini Flash models free · up to 1M context · daily request caps
- Base URL: https://generativelanguage.googleapis.com/v1beta/openai
- Suggested model: gemini-2.5-flash
- Outside the EU, UK and EEA, Google may use free-tier prompts to improve its models.
Groq
Very fast open models · ~30 req/min, ~1,000 req/day
- Base URL: https://api.groq.com/openai/v1
- Suggested model: llama-3.3-70b-versatile
Cerebras
~1M tokens/day · 8K context on the free tier
- Base URL: https://api.cerebras.ai/v1
- Suggested model: gpt-oss-120b
- The 8K free-tier context fits chat and small edits; use another provider for large agent runs.
Mistral
Experiment plan · ~1B tokens/month · includes Codestral
- Base URL: https://api.mistral.ai/v1
- Suggested model: mistral-small-latest
- The free Experiment plan requires opting in to data use for training.
OpenRouter Free
Models ending in :free · ~20 req/min, 50 req/day
- Base URL: https://openrouter.ai/api/v1
- Click “Load models”: the list is filtered to free models. A one-time $10 top-up raises the daily limit to 1,000.
NVIDIA NIM
80+ hosted models · ~40 req/min with a free developer account
- Base URL: https://integrate.api.nvidia.com/v1
- Suggested model: meta/llama-3.3-70b-instruct
Z.ai (GLM)
GLM Flash models at $0 · 128K context · ~1 req/s
- Base URL: https://api.z.ai/api/paas/v4
- Suggested model: glm-4.5-flash
Cohere
Trial key · ~1,000 calls/month
- Base URL: https://api.cohere.ai/compatibility/v1
- Suggested model: command-a-03-2025
- Trial keys are for non-commercial use.
Cloudflare Workers AI
~10,000 neurons/day free
- Base URL: https://api.cloudflare.com/client/v4/accounts/ACCOUNT_ID/ai/v1
- Suggested model: @cf/meta/llama-3.3-70b-instruct-fp8-fast
- Replace ACCOUNT_ID in the Base URL with your Cloudflare account id and use an API token with Workers AI permission.
Hugging Face
Small monthly credit on free accounts · thousands of open models
- Base URL: https://router.huggingface.co/v1
- Suggested model: openai/gpt-oss-120b
FAQ
Is a free API good enough for the agent?
For chat and small edits, yes. Long agent runs read many files, so prefer providers with a large context and daily quota, such as Gemini, Mistral or Z.ai, and switch to a paid key for heavy work.
Do free tiers use my code for training?
Some do: Google may outside the EU, UK and EEA, and Mistral's free plan requires opting in. Read each provider's terms before sending private code.