Devin (hosted) support status

Hosted Devin has no direct Future AGI integration. Learn how customer-run Devin SDK tracing differs and what content it records.

About

Hosted Devin session tracing is not supported as a direct Future AGI integration. Starting a session in Cognition’s cloud from the web app, Slack, Linear, an automation, or the Devin API does not automatically send traces to Future AGI. There is no hosted-Devin connection to configure in Future AGI.

Future AGI does not collect Devin credentials or poll Devin’s APIs. This is separate from the customer-run SDK tracing described below.

Devin SDK tracing

The Devin SDK (@cognition-ai/sdk) can record live events from both local and cloud sessions as OpenTelemetry spans. Your application runs the SDK, authenticates to Devin, and configures the tracer provider and exporter. See Cognition’s tracing guide for the native setup.

This is a customer-run SDK capability, not a supported Future AGI Devin integration or a server-side export configured in the hosted app.

Content and privacy

Content is recorded by default, including prompts, replies, and tool inputs and outputs that can contain repository code. Set recordContent: false in the SDK’s tracing options to omit that text before exporting. The setting retains timing and model/tool names, plus usage and cost when the session reports them.

You choose the export destination and manage the Devin credentials in your own application. Review custom span attributes and other instrumentation separately; the SDK’s content setting is not a general-purpose redaction filter.

Cloud-session limits

  • The SDK observes live events only while your process is attached. History replayed on attach or reload is not exported; an in-progress turn is traced from the time you attach.
  • Each turn is its own trace unless it has a parent span. Use a parent span or the active OpenTelemetry context to nest turns under an existing trace.
  • Cloud sessions report turns and tool calls, but not per-request token usage or cost. Do not expect model-request generation spans for cloud sessions.

Other Devin options

Next Steps

Was this page helpful?

Questions & Discussion