Eden AI

The Challenge

Many technical support requests required access to information that only developers could retrieve. When a user reported an incorrect balance, a failed API call, an unclear error message, or inconsistent provider behavior, the support team often needed help from engineering to investigate the issue.

The process typically looked like this:

1. A user opened a support ticket.
2. Support reviewed the request but lacked access to the necessary internal data.
3. The issue was escalated to a developer.
4. The developer interrupted their current work to inspect logs, databases, or source code.
5. The findings were sent back to support.
6. Support reformulated the technical explanation for the user.

This circuit slowed down response times and repeatedly pulled developers away from product work.

Many of these investigations were relatively simple once the right systems were accessible: confirming a credit balance, tracing an API request, identifying a provider error, or explaining the behavior of a particular endpoint.

The problem was not a lack of expertise. Support knew which questions to ask, but not always where to find the answers. Developers knew where to look, but became a bottleneck.

The Solution

Eden AI deployed an internal technical support assistant built with Claude and connected it directly to the systems used by its developers.

The assistant acts as a bridge between support and engineering. It can investigate objectively verifiable technical issues independently and escalate only the cases that require human judgment, deeper investigation, or a code change.

Through MCP connections, the assistant can access:

• Metabase to verify credit balances, consumption history, and payments
• Sentry and AWS CloudWatch to trace errors and inspect application logs
• GitHub to read and analyze Eden AI’s source code
• Linear to identify known bugs and related engineering tickets
• Intercom to understand the context of customer conversations

The assistant does more than retrieve information from dashboards. Its read access to the codebase allows it to follow the path of an API call, understand how a feature behaves, and provide precise explanations based on the actual implementation.

Persistent memory also allows the assistant to retain recurring error patterns, internal conventions, and investigation methods. Once a technical issue has been resolved, that knowledge can be reused by the entire support team.

The Results

Since deploying the assistant, Eden AI has shortened the path between a customer question and a reliable technical answer:

• Faster ticket resolution: Routine technical requests can be resolved in minutes instead of several hours
• Fewer developer interruptions: Support no longer needs engineering assistance for every balance check, log investigation, or API error
• More autonomous support: The support team can access precise, data-backed explanations directly
• More accurate answers: Responses are based on internal data, application logs, documentation, and the real source code
• Better escalation: Developers are involved only when an issue requires a code change or genuine technical judgment
• Reusable technical knowledge: Investigation patterns are remembered and shared across the team
• More product focus: Developers and technical team members can spend more time on integrations, content, code examples, and complex bugs