# usepod-agent
Provider agent for the [Use Pod](https://usepod.ai) AI inference marketplace. Operators run this on their hardware to receive dispatched inference jobs from the Use Pod coordinator and serve them via a local backend (vLLM, llama.cpp, LM Studio, Ollama) or upstream proxy (OpenRouter, Venice).
## Install
```sh
```
Linux x86_64/arm64, macOS Apple Silicon, and Windows x86_64 binaries are attached to each [GitHub Release](https://github.com/Sortis-AI/usepod-agent/releases) with SHA-256 sums.
## Configure
Copy `agent.example.toml` to `~/.config/usepod-agent/agent.toml` and fill in:
- Coordinator URL (default `wss://api.usepod.ai/provider/connect`)
- Operator wallet (Solana, where USDC payouts arrive)
- Per-model pricing
- Backend list
Then run:
```sh
usepod-agent run
```
The agent generates an Ed25519 identity key on first run (`~/.usepod-agent/identity.key`, mode `0600`), connects outbound over WSS, registers capabilities, and begins receiving jobs.
## Build from source
```sh
git clone https://github.com/Sortis-AI/usepod-agent
cd usepod-agent
cargo build --release
./target/release/usepod-agent --help
```
## Claude Code skill
A first-party Claude Code skill ships at `.claude/skills/usepod-agent-setup/`. If you run Claude Code on the host machine, it can install, pair, and service-ify the agent for you. To make it available globally:
```sh
mkdir -p ~/.claude/skills
cp -r .claude/skills/usepod-agent-setup ~/.claude/skills/
```
Then in any Claude Code session: "set up usepod-agent on this machine".
## Cross-repo
The Use Pod coordinator (the marketplace itself, USDC settlement, dashboard, SDKs) is developed in a separate, private monorepo. The wire protocol between agent and coordinator is documented in `src/ws_client.rs` (agent side) and mirrored on the coordinator side.
## License
Apache-2.0