# Surogate

> Surogate is a platform to build, deploy, and monetize
> autonomous AI agents that run in the cloud 24/7 — and to train small "expert
> models" you own that match or beat frontier models on your tasks at a fraction
> of the cost. No code required to build an agent; full SFT/RL training stack for
> ML teams.

## Product
- [Home](https://surogate.ai/): Autonomous cloud agents and expert models you own.
- [Pricing](https://surogate.ai/pricing/): Free, Standard $19/mo, Pro $96/mo, Max $190/mo, Enterprise; token allowances, compute credits, wallets, top-up rates.
- [Documentation](https://docs.surogate.ai): Agent runtime, hub, and developer toolkit.

## How it works
- Agent Builder: define an agent's goals, tools, skills, and MCP servers in plain language — no Python/ML background.
- Cloud Execution: agents run continuously in the cloud, independent of your machine.
- Expert Models (SLMs): train small specialized models on your data; SFT and RL (LoRA, full fine-tunes, GRPO, DPO).
- Bring Your Own Compute: AWS, GCP, Azure, Nebius, Lambda, RunPod, Modal, or on-prem (via dstack/skypilot).

## Pricing facts
- Free: 5 agents, 1M starter tokens, 1 GB workspace, 1 GB hub, 30 min web browsing, no card required.
- Standard ($19/mo, $18/mo annual): 100 agents, 10M tokens/mo, 5 GB workspace, 10 GB hub, 5 hours web browsing, 180 compute credits/mo (3 GPU-hours).
- Pro ($96/mo, $90/mo annual): 500 agents, 80M tokens/mo, 20 GB workspace, 50 GB hub, 15 hours web browsing, 900 compute credits/mo (15 GPU-hours), agent monetization.
- Max ($190/mo, $180/mo annual): 1,000 agents, 200M tokens/mo, 50 GB workspace, 200 GB hub, 40 hours web browsing, 2,400 compute credits/mo (40 GPU-hours).
- Compute credits meter GPU time (fine-tuning, dataset generation, evals). One credit is one minute on the baseline GPU tier; faster GPUs draw proportionally more (an H100 minute is about 6.7 credits). Training on your own connected cloud consumes no credits.
- Token top-up: $2.00 / 1M tokens (rolls over). Compute credit top-up: $2.00 / GPU-hour. Web browsing top-up: $0.08 / hour (10-hour minimum). Hub and workspace overage: $0.0605 / GB / month.
- Bring your own LLM (OpenRouter, OpenAI, Anthropic, Together, Groq, vLLM, TGI, Ollama) to bypass bundled tokens.

## Open source
- [Surogate Trainer](https://github.com/invergent-ai/surogate): training engine.
- [Agent harness](https://github.com/invergent-ai/surogates): agent runtime.

## Contact
- [Discord](https://discord.gg/eVTq93BrQ4)
- [Support](mailto:hello@surogate.ai)