Skip to navigation
Overview

Guides

Task-shaped walkthroughs for things you build on top of a deployed agent

View as Markdown

Each guide takes one goal, states what it assumes you already have, and ends with a working result. For what a thing is rather than how to do it, start from The agent lifecycle and follow it into the platform pages.

Agent capabilities

Serve a frontend from your agent

Ship your own web UI instead of the chat sidecar.

Store data with SQLite

Give your agent a place to save data that outlives a restart.

Run code in a sandbox

Let your agent write and run code on a machine of its own.

Use a user's connections

Let your agent act as the person chatting, through their GitHub, Postman, or Box account.

Using knowledge stores

Give your agent a database it can read from and write to.

Host Postgres on EKS for PrivateLink

Run Postgres with pgvector on EKS and connect it to Astropods privately.

Connect to OAuth-protected MCP servers

Complete the OAuth flow at deploy time using the agent’s public URL.

Security and access

Manually authorize requests

Reproduce the access checks the messaging sidecar runs for you.

Egress IP allowlisting

Reach a database behind your firewall over the public network.

Govern agent access with Lumos

Hand audience membership to your approval and access-review policy, with the connector that carries it.

Operations

Troubleshooting stuck deployments

Why a deploy stalls in the deploying state, and how to recover it.

Next steps