# AI Engineer WOWs at Canopy in San Francisco

> Source: <https://blog.postman.com/ai-engineer-wows-at-canopy-in-san-francisco/>
> Published: 2026-06-26 15:00:12+00:00

# AI Engineer WOWs at Canopy in San Francisco

Last week at Postman’s [Agents and APIs meetup](https://www.postman.com/events/) at Canopy in San Francisco, I gave my first external demo of the [AI Engineer](https://blog.postman.com/introducing-the-ai-engineer/) to a packed room of about 80 developers. Did I fly from LA to SF just to demo this because I was so excited? Yes. Was it worth it? Hell yes.

The headline of my talk was: *Coding is easy. Context is not.*

The thesis: software doesn’t sit in isolated files. A schema change in one place ripples through APIs, services, data flows, and teams that the AI has no visibility into. When you’re making system-level changes, generating the code is rarely the bottleneck. The bottleneck is understanding the impact of that code across the entire system.

That’s what the AI Engineer is built to solve, because it queries the [Context Graph](https://blog.postman.com/new-postman-is-here/).

## The setup: one small schema change

The demo started with a tiny edit. I changed the response schema of one API. New field, slightly different shape.

Then I prompted the AI Engineer:

@Postman Tell me what needs to be updated because of that change. What are the

[downstream dependencies]?

A couple of minutes later, the AI Engineer surfaced 10 artifacts that needed updates:

- 5 other APIs that called the updated endpoint
- The backend
`models.py`

file - The main Flask
`app.py`

file - The tests
- The docs

One schema edit. Ten places to touch. I would have never known to open most of those files on my own. The room got quiet.

I then pointed out the obvious. If one small change touches 10 artifacts, imagine a migration. Or an authentication rewrite. Or a complete refactor. Think about how many artifacts and how much of your application would silently break that you’d never know to update. That’s the failure mode every senior engineer has lived through, and it’s the failure mode that [retrieval-augmented coding agents have not solved](https://arxiv.org/abs/2308.03296) on their own. Until now.

## Why the Context Graph matters

The Context Graph creates a map across artifacts so the AI Engineer has a continuously updated understanding of how your systems are connected. APIs are the contracts that tie those systems together. By making those relationships visible and traversable, the AI can reason across the architecture instead of starting from scratch every prompt.

It can see that this endpoint belongs to this service, that this contract is consumed by these applications, and that these changes may affect these workflows. I could feel everyone leaning in.

Engineering at scale is graph traversal. The senior engineers who seem to instinctively know which Slack channel to ping before a refactor aren’t doing magic. They’re walking an internal map of producers, consumers, owners, and tests. The Context Graph encodes that map and hands it to the agent.

## Audience question 1: what’s actually inside the Context Graph?

There was time for two questions. The first perfectly summed up the point I was trying to make.

What’s inside the Context Graph?

The short answer: EVERYTHING.

- Your APIs and their
[OpenAPI specifications](https://learning.postman.com/docs/design-apis/specifications/overview/) - Your databases and services
- Your
[Postman Collections](https://learning.postman.com/docs/collections/collections-overview/), monitors, and environments - Third-party tools like Sentry, Datadog, and New Relic
- Any service that calls your APIs, whether it lives in the same repo or a different one.

Your full API ecosystem shows up in the Context Graph. That makes it discoverable, queryable, and actionable, which is the prerequisite for any agent that wants to make changes you can trust.

## Audience question 2: why invoke from Slack first?

The second question was about invocation surface.

Why Slack?

We brought the agent to where you already work, and more invocation points are coming. You can already drive it through [Agent Tasks in the Postman workbench.](https://learning.postman.com/docs/postman-ai-agent-builder/agent-mode/) The API, Claude, GitHub, and [the Postman CLI](https://learning.postman.com/docs/postman-cli/postman-cli-overview/) are on the way.

The pattern I keep coming back to: the agent should run where the work already happens. Forcing developers into a new tab is the failure mode of every internal platform tool that ever shipped.

## What this changes

The key takeaway developers walked away with: AI-assisted development falls apart at the architecture level without context, and the Context Graph is what makes it work.

When the AI engineer gave back a huge list of downstream dependencies – five other APIs, the models file, the Flask backend, the tests, the docs, you could see the excitement. It wasn’t a code generation demo. It was a system-level reasoning demo.

That’s the shift I want every developer to feel firsthand.

## Try it yourself

The AI Engineer is currently in beta. If you have a Postman workspace with collections and specs already in it, you already have the substrate. Add the integration for Slack, then ask the agent to do something that would normally require a manual audit:

```
@Postman A teammate is about to rename the `user_id` field in the
GET /users/:id response. Trace every collection, service, test, and
doc that consumes it, and tell me what would break.
```

Or, if you want to start with something safer:

```
@Postman Review the dependencies across the services in this workspace.
Surface any breaking inconsistencies and propose changes.
```

Then watch what comes back and try to pick up your jaw off the floor.

## Resources

[Introducing the AI Engineer (Postman Blog)](https://blog.postman.com/introducing-the-ai-engineer/)[The New Postman is Here: AI-Native and Built for the Agentic Era](https://blog.postman.com/new-postman-is-here/)[Postman Docs: API specifications overview](https://learning.postman.com/docs/design-apis/specifications/overview/)[Postman Docs: the Postman CLI](https://learning.postman.com/docs/postman-cli/postman-cli-overview/)[Postman Docs: integration for Slack](https://learning.postman.com/docs/integrations/available-integrations/slack/)[Postman events calendar](https://www.postman.com/events/)
