Give OpenClaw agents Claude, DeepSeek, Gemini, GPT-5, and 100+ models through one SkillBoss API key.
OpenClaw already gives you agent orchestration and skills. SkillBoss turns model access into one stable layer underneath it, so agent builders can mix providers without per-provider plumbing.
OpenClaw is an open-source AI agent platform for building, running, and managing AI agents. It provides a skill system, tool integrations, and agent orchestration — all using standard OpenAI-compatible APIs under the hood.
Package AI capabilities as skills that agents can discover and invoke. Skills define what models to use, what tools to call, and how to respond.
Run multiple agents in coordination, passing context and tasks between them. OpenClaw handles the plumbing.
OpenClaw uses standard OpenAI-compatible APIs internally, making it trivial to point agents at SkillBoss for access to 100+ models.
Your agents deserve the best model for each task — not just whichever model has the cheapest subscription.
Claude 4.5 Sonnet, DeepSeek V3, Gemini 2.5 Flash, GPT-5, Llama 4, Mistral — every model your agents might need, through one SkillBoss API key.
Agent doing quick retrieval? DeepSeek at $0.27/M. Agent reasoning through a complex plan? Claude at $3/M. Right model, right cost.
Agents make a lot of API calls. SkillBoss charges 0% markup — you pay only provider cost. Scale without paying a SkillBoss tax.
SkillBoss is fully OpenAI-compatible: tool calling, function calling, streaming, vision. Everything OpenClaw uses is supported.
2 minutes. No code changes required.
Sign up at skillboss.co/console and copy your API key. Free credits included.
Set these two env vars in your agent environment or shell:
Set the model in your agent config or skill YAML:
# skill.yaml
name: my-research-skill
model: deepseek/deepseek-v3 # cheap + fast for research tasks
provider:
type: openai-compatible
base_url: ${OPENAI_BASE_URL}
api_key: ${OPENAI_API_KEY}
# For complex reasoning, use a stronger model:
# model: bedrock/claude-4-5-sonnetexport OPENAI_BASE_URL="https://api.skillboss.co/v1" export OPENAI_API_KEY="sk_sb_xxxxxxxxxxxxxxxxxxxx" # Now run any OpenClaw agent — it will use SkillBoss openclaw run my-agent
Match the model to the task. Agents often run hundreds of calls — model choice matters.
| Agent Task | Model | Price | Why |
|---|---|---|---|
| 🔍 Research & Retrieval | deepseek/deepseek-v3 | $0.27/M tokens | Cheapest capable model, ideal for high-volume agent loops |
| 🧠 Complex Reasoning | bedrock/claude-4-5-sonnet | $3/M tokens | Best for multi-step planning, tool use, long-horizon tasks |
| ⚡ Real-time Agents | gemini/gemini-2.5-flash | $0.075/M tokens | Fastest response, ideal for interactive/real-time agents |
| 🔬 Advanced Coding | openai/gpt-5 | $10/M tokens | State-of-the-art for code generation agents |
Set OPENAI_BASE_URL=https://api.skillboss.co/v1 and OPENAI_API_KEY=your-skillboss-key. OpenClaw agents that use OpenAI-compatible APIs will automatically route through SkillBoss.
100+ models: Claude 4.5 Sonnet, DeepSeek V3 ($0.27/M — best for budget agents), Gemini 2.5 Flash (fastest for real-time), GPT-5, Llama 4, Mistral Large, and dozens more.
Yes. SkillBoss supports tool calling, function calling, streaming, and vision — all features AI agents depend on. And zero markup means agent costs scale cleanly with usage.
Yes. Each agent or skill YAML can specify its own model. Mix cheap fast models for retrieval agents and powerful models for reasoning agents, all under one SkillBoss API key.
600+ APIs, 1300+ skills — the complete toolkit for AI agents
AI Agents
Tell your agent:
set up skillboss.co/skill.mdAuto-configures base URL, auth, and model access. Works with OpenClaw and any MCP-compatible agent.
Developers
1. Get API key at skillboss.co/console
2. POST to api.skillboss.co/v1/run
3. Pick a model from 600+ APIs
Pay-as-you-go. $2 free credit. No subscription required.