Scheduled jobs (crons)
Tracks whether imports, syncs, and other scheduled work actually runs. Stuck jobs surface immediately, including history and run duration.
Monitoring that handles routine tasks on its own
LQDeck is an operations platform for companies running their own web applications. It watches scheduled jobs, errors, and site availability in one place. When code needs to change, the AI orchestrator takes the task, prepares a fix, and delivers a pull request. Your team decides only where human judgment is required.
Task
Code
PR
Merge
Agent
Implemented. Fixed null reference in import, added guard and test. PR #318 opened, CI green.
PR ready
Safe change. Waiting for your merge or auto-merge per project settings.
What LQDeck does
Connects application operations monitoring with autonomous task resolution. Connect your Nette apps through a lightweight connector, without overhauling how you develop.
Application errors
Collects production errors, groups similar occurrences, and filters noise. Your team sees what's new and what needs attention, in a clear admin.
Website availability
Checks that sites and APIs are up. On outage, sends repeated Slack alerts until service recovers, including downtime.
AI orchestrator
When a task needs code changes, the agent takes over: reads context, writes a fix, opens a pull request, and can merge safe changes on its own.
A task arrives on its own. The fix leaves as a pull request
The agent works in an isolated copy of your repository, reads code and operational data, and delivers a normal PR with green CI. You approve only what needs a human decision.
Less routine work
Small bugfixes and recurring changes don't have to wait for a free slot on the dev team.
Runs continuously
A task filed in the evening can have a proposed fix by morning, without manual triggering.
Safe changes automatically
The system scores change risk. Lower-risk work can merge automatically; the rest goes to review.
Full transparency
Every step, agent conversation, diff, and cost is visible in admin. Nothing happens in the dark.
The orchestrator doesn't replace engineers. On clear scope it handles routine; on large or ambiguous work it hands off to your team with prepared context.
How one task flows through
Takes the task
From a ticket, email, or monitoring alert.
Prepares environment
Clones the repo and works on its own git branch.
Writes the fix
Reviews code, logs, errors, and data within limits you set.
Verifies & opens PR
Security review, green CI, work and cost estimate.
Tasks stay open. A Freelo comment or email reply wakes the agent. For selected projects, a test-server check can run before the PR opens.
Where tasks come from
Work can arrive wherever your team already handles it, or directly from monitoring when something fails.
Tickets
Freelo and Asana, including attachments. Edits to the ticket re-trigger the agent.
Write to a dedicated address and a task is created. Reply in the thread to continue with the agent.
Monitoring alert
A captured error, failed cron, or outage can open a task directly.
Security is built on rules, not trust
Works on a copy
Never touches production DB or live runtime. Database access is read-only through allowed sources.
Automatic only when safe
Risky or sensitive changes always need a human. Kill switch stops the whole orchestrator.
Auditable trail
Every step, diff, and cost in admin. Slack when a PR is ready or a decision is needed.
Self-healing
Stuck runs are detected, cleaned up, and escalated to a human when needed.
Connects to your systems
LQDeck speaks HTTP API: apps report status, the agent reads context, your team connects through documented endpoints. Technical docs at /docs/api.
/api/connector
Job lifecycle from your Nette runner: schedule, start, progress, finish, fail.
/api/context
Context API for the agent: tasks, errors, job-runs, logs, DB queries.
/api/orchestrator
Worker contract for sessions, turns, messages, and global pause.
/api/…
Read-side REST for projects, crons, jobs, and errors. Project API key authentication.
Want to see how this would work for you?
We'll walk through the platform on a real project, from monitoring to an orchestrator that delivers a pull request from a task.
Already have access? Sign in to admin