Connect signals
Signals are the production and deploy data self-driving reads: errors, analytics, session replay, observability, and deployments. Connect them so the signal scout has something to look at.
Why connect signals
The signal scout reads connected providers and proposes work backed by what it finds there: an error that started after a deploy, a usage drop, a failing check. With nothing connected, the signal scout has no source, and the planner falls back to the code scouts alone.
Connecting a provider does not change what self-driving is allowed to do. It gives the signal scout read access to one more source of evidence.
Providers you can connect today
Seven providers connect today:
- Sentry for error tracking.
- PostHog for product analytics.
- Honeycomb for observability.
- Datadog for infrastructure, logs, and APM.
- Vercel for hosting and deployments.
- LangWatch for LLM tracing.
- FullStory for session replay.
Each connects read-only. Sentry connects through OAuth; Honeycomb connects with a configuration key. The others connect from their own settings card.
Providers Coldtea recognizes but does not connect yet
When the codebase mapper reads your repository, it can recognize dozens more providers than the seven above, across error tracking, analytics, feature flags, alerting, uptime, logging, and more. Recognizing a provider is not the same as connecting to it. Those extra providers are planned, not working: Coldtea can tell you it saw one in the code, but it cannot read from it yet.
If self-driving finds a provider it recognizes but cannot connect, it surfaces it as a request rather than a live connection.
How connecting works
The connection flow lives behind the Monitoring screen. Connecting a provider follows the same shape:
- Open the provider you want to connect.
- Complete its authorization: an OAuth handoff for providers that support it, or pasting a read-only key for the ones that use a key.
- Return to Coldtea and wait for the connection to show as connected.
Coldtea also surfaces providers it detected in your codebase, so the ones you actually use are easy to find rather than buried in a long list.
After connecting
Once a provider is connected, the signal scout can read it on the next daily run, and its findings show up as tasks with the source signals attached. Connect a provider before a run if you want that run to see it.
For the provider list and how these connections feed self-driving, see observability integrations.