Welcome to Astro AI
Deploy and run AI agents with agent-native infrastructure
Get started
Welcome to Astro AI
Astro AI is a platform for deploying and running AI agents. It provides agent-native infrastructure including models, knowledge bases, tool integrations, and observability. Agents are defined with a declarative spec, built and pushed with the CLI, and deployed through the API or dashboard.
Start here
How a project becomes a running agent, and which page covers each stage
Download and install the ast CLI for your platform
Create an agent, configure it, and run it locally in three steps
Build and push your agent to the registry with ast blueprint push
Run a blueprint as a live hosted agent with ast blueprint deploy
Explore
Call managed models through an OpenAI-compatible API, with no provider key
Task walkthroughs: frontends, storage, MCP servers, authorization
Connect your agent to Slack, web chat, and other platforms over gRPC
Every ast command, grouped by what it operates on
The REST API behind the CLI and the dashboard
Connect a database once and share it across your agents
Trace token usage, tool calls, and latency from any framework
Concepts
- Project: A local agent codebase with an
astropods.ymlspec. Useast project createto create an agent harness,ast project configureto set credentials, andast project startto run it locally. - Astropods Spec: Declarative YAML format (
astropods.yml) that describes an agent’s topology: container image, models, knowledge stores, tools, and ingestion pipelines. See Astropods Spec. - Blueprint: A versioned snapshot of a project pushed to the Astropods registry. Created with
ast blueprint push <name>. The same project can produce multiple blueprints under different names (e.g. staging vs. production). - Agent: A live, running instance of a blueprint deployed on the platform. Created with
ast blueprint deploy <name>. Managed withast agentcommands. - Account: The scope for all platform resources. Blueprints, agents, and secrets belong to an account. Use
ast account switchto work across personal and org accounts. - Secrets: Encrypted credentials stored in the account vault with
ast secrets. Referenced at deploy time asKEY=@SECRET_NAME. - Agent Card: A Markdown file with YAML frontmatter (
AGENT.md) that controls how a blueprint appears in the catalog: description, tags, authors, and integrations. See Make your agent discoverable.