AI operations copilot for small teams

Know why your stack is failing before you touch it.

JM Labs is building an evidence-first operations copilot for teams running self-hosted infrastructure. It connects system state, logs, and runbooks, then uses Claude to produce grounded diagnoses and safe remediation plans.

  • Read-only first
  • Evidence-backed
  • Self-hosted aware
Illustrative output

Incident analysis

82% confidence
Container Restart loop
Storage 94% used
Backup 31h stale
Likely cause

Storage pressure likely interrupted the nightly backup and triggered the application restart loop.

host.disk.root backup.nightly.latest container.app.logs
  1. Verify the backup target is reachable.
  2. Review inactive images before reclaiming space.
  3. Retry the backup and confirm the next health check.

No changes are run automatically.

The product

From scattered signals to one defensible answer.

Small teams already have monitoring. What they lack is the time to correlate every signal, check every runbook, and decide on the safest next action.

01

Collect the facts

Pull read-only context from Docker, Coolify, Airflow, host metrics, backup jobs, and existing runbooks.

02

Reason across systems

Use Claude to connect structured state with unstructured logs and operational knowledge—not just summarize alerts.

03

Keep humans in control

Cite the evidence, expose uncertainty, and propose reversible steps before any operational change.

How it works

A narrow, trust-first workflow.

  1. ConnectRead-only adapters
  2. ObserveState, logs, runbooks
  3. ReasonClaude analysis
  4. DecideEvidence + safe plan

Build status

Starting with a real self-hosted stack.

The first prototype is intentionally narrow: one operator, one stack, and one high-quality diagnostic loop. No invented traction. No autonomous production changes.

Now

Read-only prototype

Coolify and Docker health, logs, storage, and backup freshness.

Next

Repeatable incident reports

Evidence citations, confidence, and validated remediation checklists.

Later

Small-team pilots

Private deployments, shared runbooks, and human-approved actions.

Origin

Built from operating experience.

JM Labs is an independent, pre-incorporation project focused on practical operations tooling for small technical teams.

It starts from direct experience operating self-hosted infrastructure and data systems without a dedicated SRE team.

Currently building

Operations guidance that shows its work.

Product validation is being prepared.