# FlyMy.AI

> FlyMy.AI is a cloud platform for two complementary products: **Agents** (autonomous AI workflows you build, freeze, and re-run with fresh inputs) and a **Models API** (fast, scalable inference for popular open-source neural networks). This file lets AI assistants discover the documentation structure and load relevant pages on demand.

## Two distinct endpoints

- **Agents API** lives at `https://backend.flymy.ai/api/v1/agents/...` — used by the `AgentClient` from the `flymyai` Python SDK.
- **Models API** lives at `https://api.flymy.ai/api/v1/...` — used by `flymyai.run` / `FlyMyAI` from the same SDK.

API keys are issued in the user profile at https://app.flymy.ai/profile — the same key works for both endpoints (sent as `X-API-KEY` header).

## Agents

- [Overview](https://docs.flymy.ai/agents/overview): What FlyMy.AI Agents are, the agent lifecycle (create → run/chat → freeze → run instruction), and where they fit.
- [Quickstart](https://docs.flymy.ai/agents/quickstart): Five-minute end-to-end setup — create an agent, run it, get a result.
- [Agents (concept)](https://docs.flymy.ai/agents/concepts/agents): Top-level agent configuration: name, goal, status, tools, schemas.
- [Inputs, Outputs & Variables](https://docs.flymy.ai/agents/concepts/inputs): JSON Schemas (`input_schema` / `output_schema`) and plain-text descriptions (`input_description` / `output_description`) — the authoritative scope used by freeze.
- [Runs & Freeze](https://docs.flymy.ai/agents/concepts/runs): Run lifecycle states, streaming events, follow-up messages, freezing a run into a Markdown instruction, re-running with fresh variables.
- [Tools & MCP](https://docs.flymy.ai/agents/concepts/tools): Catalog of MCP tools, how to configure them, attach to agents.
- [Python SDK](https://docs.flymy.ai/agents/sdk/python): Full reference for `AgentClient` — `client.agents`, `client.runs`, `client.tools`, `client.compilations`. Covers create, list, update, delete, run, freeze, `run_instruction_and_wait`, schema suggestion, manual `instruction_md` editing.
- [CLI](https://docs.flymy.ai/agents/sdk/cli): `flymyai` CLI for agents.
- [Training with Patches](https://docs.flymy.ai/agents/guides/training): Iteratively refine an agent's goal and configuration.
- [AI assistants integration](https://docs.flymy.ai/agents/guides/ai-assistants): How to plug FlyMy.AI docs into Claude Code / Workspaces / other AI tools.

## Models API

- [Models Quickstart](https://docs.flymy.ai/quickstart): Run a single neural network in three lines (`flymyai.run(apikey=..., model=..., payload=...)`).
- [Authentication](https://docs.flymy.ai/authentication): How API keys work and where to get one.
- [Python SDK](https://docs.flymy.ai/clients/python): `FlyMyAI` sync and async clients, batching, streaming, error handling.
- [JavaScript SDK](https://docs.flymy.ai/clients/js): TypeScript-friendly client.
- [Generate Images](https://docs.flymy.ai/guides/generate-images): SDXL Turbo, Flux, nano-banana, Imagen.
- [Generate Videos](https://docs.flymy.ai/guides/generate-videos): Wan, ByteDance Seedance, Veo, JOOM.
- [LoRA Training](https://docs.flymy.ai/guides/lora-training): Train a LoRA adapter and deploy it.
- [Flux](https://docs.flymy.ai/guides/flux): Flux family models in one place.
- [Errors](https://docs.flymy.ai/errors): Error taxonomy and retry guidance.

## MCP Integrations

- [Overview](https://docs.flymy.ai/mcp/overview): What MCP is and how FlyMy.AI agents consume it.
- [Tool Catalog](https://docs.flymy.ai/mcp/catalog): Built-in MCP tools available out of the box.

## Where to start

If you're a developer who wants to **run a single neural network**, go to [Models Quickstart](https://docs.flymy.ai/quickstart) and copy the three-line example.

If you're building an **autonomous workflow** that talks to APIs, queries data, posts to Slack, etc., go to [Agents Overview](https://docs.flymy.ai/agents/overview) and follow the lifecycle: create → chat → freeze → run instruction with `variables`.

API key management lives at https://app.flymy.ai/profile.