Back to news

GitHub Copilot CLI Gains Custom Agents; Vercel Adds Claude 3.5 to AI Gateway

Two developer tooling updates this week expand how teams can automate workflows and access frontier AI models at the infrastructure layer.

GitHub Copilot CLI Gains Custom Agents; Vercel Adds Claude 3.5 to AI Gateway

What happened

GitHub has introduced custom agents for Copilot CLI, enabling engineering teams to encode their specific stack configurations and internal workflows directly into the terminal assistant, transforming ad-hoc AI prompts into structured, repeatable processes. Rather than relying on generic one-shot queries, teams can now define agents that understand project context and produce outputs that can be reviewed and reused across the organization. Separately, Vercel has added Anthropic's Claude 3.5 model to its AI Gateway product, broadening the roster of frontier models accessible through a single unified routing layer.

Why it matters for your business

For engineering leaders, the GitHub update addresses a persistent gap: AI assistance in the terminal has largely been informal and inconsistent, meaning different developers generate different results for the same task. Custom agents impose structure without sacrificing flexibility, making AI-assisted operations more auditable and easier to onboard new engineers into. On the Vercel side, adding Claude 3.5 to AI Gateway means product teams can switch or compare models without rewriting integration code, reducing vendor lock-in and letting organizations respond faster when a new model better fits a use case. The practical takeaway is that both updates push AI tooling from experimental to operational — the kind of reliability shift that justifies broader internal adoption.

What to watch next

The trajectory of GitHub Copilot CLI suggests agent definitions will become a form of institutional knowledge — teams that invest early in building well-scoped agents will compound productivity gains as the tooling matures. On the model-access front, Vercel's gateway strategy mirrors what cloud providers have done with compute abstraction, and further model additions are likely as competition among frontier labs intensifies. Organizations evaluating AI infrastructure should track whether gateway-style abstraction layers become the de facto standard for enterprise model consumption.

Sources

Want this kind of clarity applied to your own systems?

HashWhales can review your website, infrastructure, security posture, and growth bottlenecks, then send a prioritized action plan.

Free AuditChat on WhatsApp