AI Agents
Build autonomous agents that pay for their own compute and inference — no API keys, no billing dashboards. Your agent holds a wallet, calls LibertAI models via x402, and manages its own Aleph Cloud credits to stay alive.
When to use this
The LibertAI agent stack is a thin layer over Coinbase AgentKit, not a full framework. Pick it when you want:
- Wallet-funded inference — the agent pays per call in USDC, no one issues it an API key.
- Onchain interaction — you need a wallet, tool calling, and EVM action primitives in one bundle.
- Self-hosted runtime — you ship a Docker image to your own Aleph VM with one command.
If you mainly want orchestration (memory, planning, tracing) on top of an OpenAI-compatible endpoint, a framework like LangGraph or CrewAI pointed at https://api.libertai.io/v1 with a regular API key is a simpler path. See Quickstart. The two approaches compose — you can run a LangGraph agent inside a wallet-funded loop.
How it works
- Wallet — the agent has a private key with USDC on Base
- Inference — each LLM call is paid individually via x402 (no API key needed)
- Compute — the agent monitors its Aleph credit balance and buys more when needed
- Actions — using Coinbase AgentKit, the agent can interact with any onchain protocol
Stack
| Layer | What | TypeScript | Python |
|---|---|---|---|
| Framework | Coinbase AgentKit | @coinbase/agentkit | coinbase-agentkit |
| Inference | LibertAI API (OpenAI-compatible) | openai | openai |
| Payment | x402 protocol | @libertai/x402 | libertai-x402 |
| Plugin | Wallet, credits, tools bridge | @libertai/agentkit-plugin | libertai-agentkit-plugin |
Get started
- Getting started — build the agent locally (TypeScript and Python parity)
- Deployment — push it to an Aleph Cloud VM with one command
See also
- x402 payments — payment protocol the wallet uses
- Text API — models with function-calling support
- Architecture — what happens after the agent calls the API
- GitHub source