Announcing our Integration with GitHub Copilot

Flightcrew now integrates with Github Copilot

Over 50,000 engineering teams use GitHub Copilot as their coach, muse and assistant. We are proudly one of them.

But Copilot’s blindspot is that it can’t see where your code runs and how it performs. Without access to Cloud APIs, alerts or observability it can’t answer questions like "Will this code change break anything" or "Why is my service failing?" In short, Copilot can't help you with the cloud.

We’re building the intelligence layer that catches and fixes the thousands of things that can go wrong in cloud infrastructure. Now everything Flightcrew knows about scaling, maintaining and securing cloud infrastructure is accessible via GitHub Copilot

Chat with Flightcrew

Wherever you talk to Copilot, you can loop in @flightcrew-copilot to ask questions about IAC, Kubernetes, or whatever is slowing you down.

Flightcrew constantly checks your cloud, code, and observability platforms for vulnerabilities, maintenance, and optimization opportunities. All recommendations are based on your own instance’s data and knowledge graph - this means Flightcrew can’t hallucinate a bad address or a non-existing service.

Start with prompts like:

• What is broken? What should I prioritize?
• How can I fix service?
• What does this file do?
• What does this config do?
• What happens if I change it to X?

Why build a Copilot Extension?

First, we’re heavy users of Copilot and and so this started as a hack for internal use.

Second, we’re building Flightcrew based on a few principles that point straight to Copilot:

• Integrate intelligence where engineers work
• Automate through native workflows
• Everyone trusts, but verifies

The first manifestation of these principles was our Github Application, which analyzes PRs for breaking changes, and generates PRs to fix and maintain infrastructure. This "Dependabot-style" pattern is perfect for high volume, low risk tasks where you only need a human in the loop to review and approve changes.

However, as we expand into lower frequency, higher risk areas like networking, and IAC refactoring we've found that users need to be coached, or coaching earlier in the workflow. Chat / Copilot interaction patterns seems to be the optimal workflow for tasks where …

• Users don’t know exactly what they want - usually a tradeoff workflow
• Users need to retrieve or provide additional context about a state, problem, or solution

We hope our Copilot Extension helps engineers solve more ambiguous problems, and uncovers additional preferences and context for our models. Stay tuned for further integrations with IDE + codegen tooling.

Install the Flightcrew Github Copilot Extension

You can install the Flightcrew Copilot Extension on the GitHub Marketplace … or from within your instance of Flightcrew. Note that our Copilot Extension requires a Flightcrew account.

Excited? Curious? Skeptical? Give us a ring: hello@flightcrew.io