pollinations.ai × OpenClaw

Use 25+ AI models as your OpenClaw brain through a single API. Free models with web search, vision & reasoning — connect your account and start building in 30 seconds.

Step 1

Get your free API key (comes with free credits):

or paste your key manually: enter.pollinations.ai

Step 2

Install OpenClaw, then run the setup script:

curl -fsSL https://openclaw.ai/install.sh | bash

Run the setup script:

curl -fsSL https://raw.githubusercontent.com/pollinations/pollinations/main/apps/openclaw/setup-pollinations.sh \ | bash -s -- YOUR_API_KEY

🔑 Saves API key to ~/.openclaw/openclaw.json
🤖 Adds 7 models: Kimi K2.5, DeepSeek, GLM, Gemini + Search, Claude Haiku + 2 paid
🔍 Configures web_search tool with Perplexity via pollinations.ai (preserves existing if already set)

Or paste this into ~/.openclaw/openclaw.json:

{ "env": { "POLLINATIONS_API_KEY": "YOUR_API_KEY" }, "agents": { "defaults": { "model": { "primary": "pollinations/kimi", "fallbacks": ["pollinations/deepseek", "pollinations/glm"] }, "models": { "pollinations/kimi": { "alias": "Kimi K2.5" }, "pollinations/deepseek": { "alias": "DeepSeek V3.2" }, "pollinations/glm": { "alias": "GLM-4.7" }, "pollinations/gemini-search": { "alias": "Gemini + Search" }, "pollinations/claude-fast": { "alias": "Claude Haiku 4.5" }, "pollinations/claude-large": { "alias": "Claude Opus 4.6" }, "pollinations/gemini-large": { "alias": "Gemini 3 Pro" } } } }, "models": { "mode": "merge", "providers": { "pollinations": { "baseUrl": "https://gen.pollinations.ai/v1", "apiKey": "${POLLINATIONS_API_KEY}", "api": "openai-completions", "models": [ { "id": "kimi", "name": "Kimi K2.5", "reasoning": true, "input": ["text","image"], "contextLength": 256000 }, { "id": "deepseek", "name": "DeepSeek V3.2", "input": ["text"], "contextLength": 128000 }, { "id": "glm", "name": "GLM-4.7", "input": ["text"], "contextLength": 128000 }, { "id": "gemini-search", "name": "Gemini + Search", "input": ["text","image"], "contextLength": 128000 }, { "id": "claude-fast", "name": "Claude Haiku 4.5", "input": ["text","image"], "contextLength": 200000 }, { "id": "claude-large", "name": "Claude Opus 4.6 (Paid)", "input": ["text","image"], "contextLength": 200000 }, { "id": "gemini-large", "name": "Gemini 3 Pro (Paid)", "reasoning": true, "input": ["text","image"], "contextLength": 1000000 } ] } } }, "tools": { "web": { "search": { "provider": "perplexity", "perplexity": { "baseUrl": "https://gen.pollinations.ai/v1", "apiKey": "${POLLINATIONS_API_KEY}", "model": "perplexity-fast" } } } }, "gateway": { "mode": "local" } }

Paste this into your agent chat — it will configure Pollinations as a provider:

Add pollinations.ai as a model provider. Run this command: curl -fsSL https://raw.githubusercontent.com/pollinations/pollinations/main/apps/openclaw/setup-pollinations.sh | bash -s -- YOUR_API_KEY The script will restart the gateway automatically. After that, switch to the new provider with: /model pollinations/gemini-search

Which model should I use?

Kimi K2.5 default

pollinations/kimi

Best open model for agentic coding. 256K context, vision, tools, reasoning. 77% SWE-Bench.

Gemini + Search web search

pollinations/gemini-search

Google Gemini with always-on web search. Vision support. Look up docs, APIs, recent changes.

DeepSeek V3.2

pollinations/deepseek

Strong reasoning & tool calling. Good alternative to Kimi for hard problems.

GLM-4.7

pollinations/glm

Z.ai — coding, reasoning, agentic workflows. Competitive with Kimi.

Claude Haiku 4.5

pollinations/claude-fast

Anthropic — fast with good reasoning. Claude quality, lower cost.

Claude Opus 4.6 paid

pollinations/claude-large

Anthropic's most capable model. Vision, tools, deep reasoning.

Gemini 3 Pro paid

pollinations/gemini-large

Google's flagship with vision, audio & video understanding.

Switch models anytime: /model pollinations/gemini-search

Image, video & audio

Community-built pollinations.ai skill by @isaacgounton — gives your agent multi-modal generation.

/skill install isaacgounton/pollinations

This adds image generation (Flux, GPT Image, Imagen 4), video (Veo 3.1, Seedance), text-to-speech (13 voices), and vision analysis to your agent. Separate from the LLM provider setup above.