an autonomous agent for SQL Server self-healing & network monitoring
A field manual for building a local AI agent that lives in front of a production
SQL Server. Why cloud LLMs are wrong for regulated database work, what the
sensor / reasoning / actuator architecture actually looks like in code, and what
eighteen months of putting one of these systems through real operational
conditions taught me. 69 pages. Free to read, free to download.
You inherit a database with no documentation, 200 tables, 400 stored procedures, and a senior DBA who left in March. The traditional onboarding takes two weeks. With an LLM in the loop, it is two days. Here is the playbook.
Threshold alerts catch the bug you predicted. Statistical methods catch the drift you didn't. LLMs add the explanation that makes the alert actually useful at 2 AM. Mixing the three buys you the kind of monitoring that catches plan regressions before they become incidents.
SQL Server's missing-index DMV recommends every index it thinks would help a query — and is right about half the time and catastrophic the other half. The AI alone is no better. The combined workflow gives you a smaller set of better-targeted index changes.
From a vague idea to a deployed automation in about the time it takes to finish your coffee.
1
Sketch the flow
Drag boxes, draw arrows. Describe the trigger in plain English.
2
AI fills the gaps
It writes the SQL, the cron, and the alert. You read it and tweak.
3
Ship & sleep
Roll it out with one click. Auto-rollback if anything looks wrong.
~ a peek at the cockpit ~
Your databases, drawn.
Charts you can actually read. Numbers you can actually trust. No 12-color heatmap nightmares.
Bob · Home — the console. Health, evolution, situation, monitor — all on one screen.Bob · SQL Health — instance metrics, agent jobs, disk volumes, CPU and memory trends.Bob · Evolution — cycles, applied mods, score trend, strategy mix, sacred functions, reasoning styles.Bob · Forge — self-modification pipeline. Propose, score, smoke-test, apply — or roll back.Bob · Soul — memory, identity, authority. The part of Bob that survives a restart.Bob · Kanban Board — every action Bob proposes lives here. Approve, reject, or watch it execute.Bob · Status Report — the automated health digest. Heartbeat, fleet, network catalog — what Bob tells you when nobody is watching.
~ how to get bob ~
Three ways to bring Bob home.
Start with the field kit and wire it up yourself, run a guided pilot in your environment, or hand the keys to a regulated estate over for a sovereign install. All prices one-time, no subscription.
Field Kit
$1,999
For the DBA who wants to wire Bob into their own home lab and learn by doing.
Full Bob source release
Reasoning-layer prompt templates and DMV probe queries
Three-layer architecture reference, ready to deploy on your hardware
Bob is not plug-and-play. Every tier includes training and calibration to your specific environment – your SQL Server workload, your wait-stat baselines, your remediation policies, your team's escalation conventions. The Field Kit ships with prompt templates and two weeks of installation Q&A to do this yourself. The Pilot reserves 90 days of guided weekly tuning. The Sovereign assigns a named on-call engineer for 12 months. The agent gets sharper the longer it runs in your stack.
A note on the free stack: Ollama and Ubuntu Server are both free and open-source. You can download them yourself from ollama.com and ubuntu.com/server at no cost. We do not charge for either. We charge for the Bob system: the agent code, the integration work, the training, and (on Pilot and Sovereign) the hardware that runs it.
All tiers ship with the open-weight model and a local Ollama install. Zero ongoing license fees. The Pilot tier includes a consumer-grade GPU host; the Sovereign tier includes a commercial-grade EPYC rack server with a datacenter GPU. The Field Kit is software-only. Power and rack space remain the buyer's responsibility on every tier. Need a different shape (multi-region, air-gapped, custom audit package)? Use the contact form.
Ready to talk shop?
Pick a tier, send a message, or ask the question that does not fit any of the boxes above. Replies come from a human inside one business day.