Run production tasks
with agents you can trust.
BitSentry hands your agents a runbook to follow, the exact steps you already approved. AI explains what it finds, and every action is logged.
Free during public beta ·
Running a team? See BitSentry Cloud
Runs on your machine Bring your own AI provider
Live integrations · see what's coming next →
Why runbooks
An agent in production
is a risk you can't audit.
Unlike a prompt, where you can't be sure what an agent will do, a runbook is the exact step-by-step procedure you define up front, so you know what it will run every time.
Here is what changes:
An agent left to improvise
An agent on a runbook
Hand an agent a prompt and you get a different result every time. Hand it a runbook and it runs the exact steps you already approved.
Trust-layer for agents
Finding issues is easy.
Trusting agents isn't.
The agent picks a runbook, but the runbook runtime is what touches your systems. It runs only the checks and fixes you approved, with every action logged.
Observability
Alert fires
Agentic AI
Sees the alert, executes a runbook
BitSentry Runbook Runtime
Runs the approved checks and fixes
Infrastructure
Execute
Codebase
Check & update
PRODUCTION IN ACTION
Watch incident investigation
runbook automation in action.
Two minutes. One real incident, diagnosed step by step in BitSentry Desktop.
Honest objection
Why not just use
Claude Code ?
Claude Code
-
Ad-hoc conversation
Every incident starts from scratch.
-
Context lost after chat
The next on-call rebuilds the same investigation.
-
Hard to share as process
Tribal knowledge stays with whoever typed the prompt.
-
Weak audit trail
Hard to reconstruct what ran and what AI said for compliance.
-
No control plane
The agent can run whatever command it decides on.
BitSentry Desktop
-
Repeatable diagnostic runbook
Build the investigation once. Run it the same way every time.
-
Investigation history retained
Every run, every output, every AI reading is captured.
-
Shared runbook library
Junior on-calls inherit senior engineers' muscle memory.
-
Evidence trail
Command, output, and AI response linked for the post-mortem.
-
An agent you can trust
It runs only the steps you approved, never whatever it decides on its own.
Bring your own key
Bring your own AI provider
No vendor lock-in.
- Claude Code
- Codex
- Kilo
- OpenCode
- OpenAI
- Anthropic
- Groq
- Gemini
- OpenRouter
- Ollama
No middleman
Nothing extra to buy and nothing sitting in between. BitSentry Desktop talks to your CLI or provider directly, and your logs never touch BitSentry servers.
Existing integrations
Works With Your Stack
Pull context from the tools you already use during incident investigation. Push findings into your workflow when needed.
SigNozPrivacy first
Everything stays
on your machine.
Everything runs locally on your machine, and AI analysis goes directly to the provider you choose.
Your Credentials Stay Local
BitSentry Desktop connects with the access you already have. BitSentry stores no keys and no credentials. They never leave your machine.
Every Action Logged
Full execution log for every runbook run. Shell output, AI responses, and HTTP calls are all saved locally for review and compliance.
Runs on Your Machine
No proxy in the middle. Connections go straight to your servers, AI calls go to the provider you choose, and BitSentry never sees your logs.
Try it for free
Get BitSentry Desktop
Free during public beta. Bring your own AI provider. Everything stays on your machine.
macOS · Apple Silicon & Intel