Understand a Clever Cloud production incident, from Claude or ChatGPT
A failed deployment, 500 errors, an app that restarts: Clever Cloud deployments, logs and metrics in your AI, with no console access.
Published on October 28, 2026
In short
- Connect Clever Cloud to your assistant through Persyst: it cross-checks an application’s deployments, logs and metrics to explain an incident.
- A product owner or support reads production logs with no console, no CLI and no token.
- Read-only: no deployment, restart or configuration change from the assistant.
Customers report payment errors since 9 am. What happened on the API?
A deployment went out at 9:08 am (commit “checkout refactor”). Since then, 112 responses with status 500, all on the payment route, with the same error in the logs: a “currency” field missing from the request to the payment provider. Memory and CPU are normal: this is not a load issue. The previous deployment, from 6 pm yesterday, showed no error on that route.
Sources consulted
- Clever Cloud · app-api · deployments
- Clever Cloud · app-api · logs since 9 am
- Clever Cloud · app-api · metrics
The problem: logs belong to whoever has the console
When a customer reports an error, support and product want to know what is going on. The answers are in the application’s deployments, logs and metrics, but only developers have access. They get interrupted to copy and paste log lines.
What the assistant does, step by step
With Persyst’s Clever Cloud connector, the question “What happened on the API since 9 am?” chains:
- Find the application among those opened to the member (
list_clever_cloud_apps). - List recent deployments: date, status, commit (
list_clever_cloud_deployments). - Read the period’s logs, filtered by level or text (
read_clever_cloud_logs). - Check the metrics: CPU, memory, requests (
read_clever_cloud_metrics), to rule out a load issue.
The assistant then ties the error to the deployment that introduced it, with the log lines that show it.
Three questions to try
- “Why did the application restart last night?”
- “How many 500 errors on the API since this morning, and on which routes?”
- “Has the worker’s memory been growing since Monday’s deployment?”
Setup
- An administrator authorizes Persyst from their Clever Cloud account and declares the applications to share (Clever Cloud guide).
- They open each application to the people concerned: product, support, engineering leadership.
- Each member adds Persyst to their assistant (getting started).
Limits to know
- Read-only: no deployment, restart, scaling or variable change.
- No access to secrets: environment variables and add-on credentials stay hidden.
- Logs go back no further than Clever Cloud’s retention.
Questions
Frequently asked questions
Does each person need a Clever Cloud account?
No. An administrator authorizes Persyst once with their account. Authorized members look at the opened applications from their assistant, with no account and no console access.
Can the assistant roll back to the previous deployment?
No. The connector is read-only. Redeploying, restarting or changing a configuration happens in the Clever Cloud console, by the people allowed to.
Are environment variables visible?
No. Environment variables and add-on credentials are not exposed to the assistant.
Can I open a single application?
Yes. Access is set per application, then per tool: deployments, logs, metrics. One team sees the API, another the website, nobody sees the billing application.