October 7, 2026

AI agents are moving beyond experiments. They answer questions, run workflows, call tools, and collaborate across real teams. Building agent logic is only the beginning. You also need a repeatable way to configure, test, deploy, and observe agents in production.

Astropods in the Postman AI stack

is a product in the Postman AI stack. Postman is building the infrastructure for agents and APIs, with each product covering a specific layer that helps customers achieve AI sovereignty. Astropods is the agent operating system layer: it runs and controls AI agents in production with runtime, observability, and guardrails at scale.

gives you that lifecycle in one platform. This guide takes you from a local project to a deployed, monitored agent.

What you will do: create an agent, run it locally, publish a blueprint, deploy it, connect Slack, and inspect production behavior.

Understand the Astropods lifecycle

organizes an agent into four stages:

The declarative astropods.yml file describes the agent’s models, interfaces, integrations, knowledge stores, and runtime requirements.

Start with the Astropods documentation or review the complete agent lifecycle.

1. Build and run an agent locally

Install the ast CLI, start Docker, and create a project.

The --model gateway option connects the project to the AI Gateway, so the agent can use managed models without a separate provider API key.

The generated project includes:

astropods.yml, the agent specification.
AGENTS.md, project guidance for coding agents such as Codex, Claude Code, Copilot, or Cursor.
CLAUDE.md, additional instructions for Claude Code.

Implement the agent manually or ask a coding agent to build the required logic. If the agent needs external credentials, configure them locally.

Astropods starts the agent and its sidecars in containers. Messaging agents can use the local chat interface at http://localhost:3100.

Follow Your first project for the current quickstart.

2. Publish and deploy a blueprint

A blueprint packages the agent’s container image and registered specification into a versioned artifact.

Validate the specification, then publish a private blueprint.

Blueprints are private by default. Use --visibility public only when you intend to publish to the public catalog.

Keep AGENTS.md current. It becomes the Agent Card shown on the blueprint page and should explain what the agent does, which integrations it needs, and how to use it.

Deploy the blueprint with an authenticated web interface.

If the deployment needs a credential from the account vault, reference the stored secret instead of exposing its value.

Verify the result with ast agent list, then open the launch URL and run a representative request.

See Your first blueprint and Deploy your first agent for the full workflow.

3. Connect the agent to Slack

Slack puts the agent where your team already works. In the Slack API dashboard:

Create an app for the target workspace.
Generate an app-level token with connections:write.
Enable Socket Mode and Event Subscriptions.
Add the required bot scopes. A common starting point is chat:write and app_mentions:read.
Install the app and copy the bot token.

Select the Slack adapter in the deployment form and provide the app and bot tokens. You can also deploy through the CLI.

Your organization may require administrator approval before you can install the Slack app. Configure access separately for the web and Slack interfaces.

4. Operate and monitor the agent

Deployment starts the operational lifecycle. Astropods provides three levels of visibility:

Monitor: inspect token usage, spend, active users, request volume, latency, and network activity.
Traces: open a specific interaction to review its input, output, model calls, tool calls, timing, cost, and observation tree.
Deployments: inspect workload status, deployment history, and runtime health.

This separation helps you distinguish agent behavior problems from infrastructure problems.

Learn more in Managing your agents and Monitor your agents.

Essential production commands

Production readiness checklist

ast spec validate succeeds.
The main workflow works locally and after deployment.
Required variables are documented and no credentials are committed.
The Agent Card explains the blueprint and its integrations.
Web and Slack access policies match the intended audience.
A production request produces a complete trace.
Token usage, latency, network activity, and logs look healthy.

Start building

Astropods gives you one workflow to define an agent, test it locally, publish a versioned blueprint, deploy it securely, and understand its behavior in production.

Ready to deploy your first agent? Follow the hands-on Postman Learning Path:

Building with Agents learning path

You can also start with Your first project or explore the Astropods documentation.