Self-host Nestor for free, forever. You bring your own LLM keys and pay providers directly at their published rates. Hosted tiers in development add multi-user features without locking down the core platform.
For developers and small teams.
$0/forever
MIT license. Run anywhere.
nestor_build_toolFor power users who want a hosted Studio.
$15/month
Coming Q3 2026 (estimated). Pricing not final.
your-name.nestor.shFor teams of 2–25.
$49/seat/month
Coming Q4 2026 (estimated). Pricing not final.
For organizations with strict compliance requirements.
Custom
Coming 2027. Talk to us early.
Every LLM call you make goes directly from your machine (or hosted instance) to your chosen provider, using your own API key. Nestor never proxies tokens or marks them up. You see your usage on the provider's dashboard exactly as you would when calling the API directly.
| Provider | Approximate cost | Notes |
|---|---|---|
| Anthropic Claude | $3 / $15 per 1M tokens (Sonnet 4.6) | Excellent tool use. Default for most agents. |
| OpenAI | $5 / $15 per 1M tokens (GPT-4o) | Strong reasoning + vision. Good fallback. |
| Google Gemini | $1.25 / $5 per 1M (2.0 Pro) | Free tier available for low volumes. |
| xAI Grok | $3 / $15 per 1M (Grok 3) | Real-time web access built in. |
| Mistral | $2 / $6 per 1M (Large) | EU-hosted. GDPR-friendly. |
| OpenRouter | Varies (300+ models) | Single key for many providers. Slight markup. |
| Ollama | $0 (electricity only) | Fully local. Llama 3, Mistral, DeepSeek, etc. |
Nestor's budget caps protect you from runaway spend: per-run and per-day USD limits, hard-cap on/off, and warning events emitted at 80% usage. See the configuration docs for details.
Yes. The nestor-sh npm package is open source under the MIT license. You can self-host it on your own machine or server at no cost. The future hosted tiers (Solo Pro, Team, Enterprise) add convenience features but never lock down the core platform — the MIT version stays fully functional.
Bring Your Own Key. You hold your own API keys for each LLM provider, pay them directly at their published rates, and Nestor never proxies through a paid gateway. This is also the only ToS-compliant way to use Anthropic Claude in third-party tools post the April 4, 2026 Consumer Terms update — Nestor never touches subscription tokens. See vs ClaudeClaw for the full compliance breakdown.
Self-host + Ollama. Run ollama pull llama3.2, set Ollama as your provider in nestor-sh install, and your LLM cost is electricity. Performance is lower than Claude Sonnet for complex tool use but fine for most workflows. Combine with Claude or GPT-4o for the hard tasks via the HotSwapAdapter.
The self-host tier is already free. When hosted tiers launch, OSS maintainers will get free Solo Pro accounts on request — no formal program yet, just email laurent@nestor.sh.
No. The self-host tier never sends data anywhere except the LLM providers you configure. Hosted tiers (when launched) will follow the same principle: your prompts, agents, and outputs stay on your tenant. We do not train models on customer data.
The nestor-sh skill translate command imports tool definitions from LangChain, CrewAI, OpenAI Function Calling, and MCP. Most agents migrate in under an hour.