Zero-trust operational intelligence
for AI agents in production
BitSentry is the trust layer between your agents and production. Agents never touch your servers directly, only the runbooks your team approved.
Data doesn't leave your cloud Bring your own AI provider Every agent action logged
app.bitsentry.ai / diagnosis
Diagnosis Records
| Rule Description | Agent | Severity | Recommendation | Actions | |||
|---|---|---|---|---|---|---|---|
| #11 | TRPCClientError: No authorization token provided — {TRPCClientError#2}.from(@trpc+client@11.10.0_@trpc+server@11.10.0_typescript@7.0.2__typescript@7.0.2/node_modules/@trpc/client/dist/TRPCClientError-apv8gw59) | sandbox-dashboard | high | Completed | Staging tRPC requests are rejected without authorization; root cause and exploitability remain unconfirmed. Identify the... | 2026-07-17 00:223 months ago | |
| #14 | Error: PostHog API 400: OFFSET is not supported on queries made with a personal API key. For pagination, use keyset pagination on the `timestamp` column, e.g. `WHERE timestamp > :last_seen_timestamp ORDER BY timestamp LIMIT N`. | desktop | high | Completed | PostHog rejects OFFSET pagination with personal keys; this indicates an API mismatch, not confirmed compromise. Replace... | 2026-07-17 10:333 months ago | |
| #23 | Error: Cannot find module './kernel' — Module._resolveFilename(loader) | sandbox-api | critical | Completed | Restore the missing ./kernel module and verify the deployed release before redeployment. Inspect the stack trace, caller... | 2026-07-20 05:573 months ago | |
| #25 | Error: Cannot find module './runbooks.schemas' — Module._resolveFilename(loader) | sandbox-api | critical | Completed | Staging startup fails because Node.js cannot resolve the relative ./runbooks.schemas module. Inspect the stack trace and... | 2026-07-20 12:513 months ago | |
| #28 | Error: Cannot find module './webauthn.use-cases' — Module._resolveFilename(loader) | sandbox-api | critical | Completed | Restore the missing WebAuthn module and verify authentication fails closed during startup failures. Inspect deployed fil... | 2026-07-20 14:333 months ago | |
| #35 | Error: listen EADDRINUSE: address already in use 0.0.0.0:3000 — Server.setupListenHandle [as _listen2](node:net) | sandbox-api | high | Completed | Port 3000 is occupied, preventing the Node.js application from starting; no vulnerability is indicated. Identify the lis... | 2026-07-24 03:283 months ago | |
| #36 | Error: ENOENT: no such file or directory, open '/app/packages/i18n/dist/locales/en-GB/emails.json' — bootstrap(main.ts) | sandbox-api | high | Completed | Staging startup fails because the required en-GB email translations file is missing. Verify generation, packaging, and d... | 2026-07-24 09:363 months ago | |
| #51 | TRPCClientError: Failed to fetch (api.sandbox.bitsentry.ai) — {TRPCClientError#2}.from(@trpc+client@11.10.0_@trpc+server@11.10.0_typescript@7.0.2__typescript@7.0.2/node_modules/@trpc/client/dist/TRPCClientError-apv8gw59) | sandbox-dashboard | high | Completed | Investigate the staging API fetch failure; telemetry confirms no root cause, exploitation, or CVE. Check API health, dep... | 2026-08-07 23:522 months ago | |
| #57 | TRPCClientError: Invalid `prisma.incidentMessage.create()` invocation: Unique constraint failed on the fields: (`"threadId"`, `"sortOrder"`) | PostHog Sandbox API EU | high | Completed | Application integrity issue from Prisma unique-constraint violation on incidentMessage(threadId, sortOrder); not a secur... | 2026-08-07 23:522 months ago | |
| #62 | PrismaClientKnownRequestError: — GET /api/worker/runbook-executions/claim-next | sandbox-api | high | Completed | Prisma query timeout on investigationSession.findMany() in sandbox-api staging; investigate slow query and DB health. Ca... | 2026-08-17 02:412 months ago | |
| #573 | Error: write EIO — afterWriteDispatched(stream_base_commons) | sandbox-api | critical | Completed | Fatal socket write I/O failure is confirmed; underlying infrastructure cause and security impact remain unverified. Corr... | 2026-09-04 08:48last month | |
| #1086 | sshd: authentication failed. | bitsentry-sandbox | - | Verification Pending | Repeated SSH password failures target root; no successful access or exploitation is currently evidenced. Review SSH logs... | 2026-09-08 07:10last month | |
| #1684 | sshd: Attempt to login using a non-existent user | bitsentry-telemetry | medium | Failed | Repeated invalid-root SSH attempts indicate probing; no successful authentication is evidenced. Review SSH logs for succ... | 2026-09-08 07:40last month | |
| #1713 | Multiple Windows Logon Failures | win-dc-01 | - | Failed | Repeated failed Windows logons suggest probing, but authentication details remain partially verified. Retrieve raw Event... | 2026-09-08 07:45last month | |
| #30820 | ZodError: [ — POST /api/worker/remediations/claim | sandbox-api | high | Completed | The endpoint rejects requests missing required idempotencyKey; handling and deployment context remain unverified. Valida... | 2026-09-16 11:524 weeks ago | |
| #58228 | Cron failure: ticket-status-sync | sandbox-api | high | Completed | A resolved cron failure is confirmed, but its cause and any security impact remain unverified. Retrieve full Sentry even... | 2026-09-23 07:373 weeks ago | |
| #64437 | Cron failure: ticket-status-sync | sandbox-api | high | Completed | The resolved Sentry alert does not establish the cause or confirm successful ticket synchronization. Retrieve the latest... | 2026-10-13 15:352 days ago | |
| #64440 | PAM: User login failed. | bitsentry-telemetry | medium | Verification Pending | Repeated PAM failures for the same account point to a credential sweep; verify the source address and lockout policy. Re... | 2026-10-13 16:022 days ago | |
| #64512 | Downtime detected for https://api.sandbox.bitsentry.ai | sandbox-api | critical | LLM Assessed | Uptime monitor reports the staging API unreachable; confirm the container restarted cleanly and that port 3000 is bound... | 2026-10-14 09:12yesterday | |
| #64519 | Error: API provider "groq" is disabled by ENABLED_API_PROVIDERS. Add "groq" to the flag to enable it. | sandbox-api | medium | LLM Assessed | A request selected a provider that the deployment flag excludes; this is a configuration mismatch, not an outage. Confirm... | 2026-10-15 02:4816 hours ago |
BitSentry Cloud on a sample workspace. Approve the proposal and follow the run.
Connect the tools already involved in your incidents · 19 available today →
Agents are already in production
Governance isn't
85% of surveyed organizations already run AI agents in production (CSA, 2026), and 65% reported at least one agent-related incident in the past year (CSA, 2026). This is what agent access without a control layer looks like.
-
01 Untracked agents
82%
82% of teams discovered agents in their environment that nobody had tracked.
Cloud Security Alliance (2026) -
02 Excess access
74%
74% report agents with more access than their tasks require.
Cloud Security Alliance (2026) -
03 Attribution gap
68%
68% cannot clearly separate agent activity from human activity.
Cloud Security Alliance (2026) -
04 Autonomous actions
8 in 10
Roughly 8 in 10 say agents have already taken consequential autonomous actions.
Kore.ai Agent Productivity Index (2026)
Self-reported survey data, not BitSentry measurements.
Agents choose the runbook
You decide what it can do
An agent that can think shouldn't automatically get to act. The agent picks the runbook, and the runtime runs it: only the checks and fixes you approved, every action logged.
Free PDF, no email required
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
Watch an agent diagnose a
production incident in real time
Two minutes. One real incident, diagnosed step by step through a BitSentry runbook.
Built for teams already
running agents in production
BitSentry is for production work that is repeated, spread across many systems, and too risky to let an agent run without an audit trail.
DevOps agencies and MSPs
Many client environments. One diagnostic playbook. Junior technicians stop escalating to senior infra people.
See the use caseLean on-call rotations
Two or three engineers carry the pager. Whoever is on call at 2am runs the same runbook the senior engineer wrote.
See the use caseSupport engineering
Customer says the import is stuck. Inspect production safely without giving everyone shell access or escalating to backend.
See the use caseRegulated teams
SOC 2, HIPAA, PCI. Every command, output, and AI response captured for the audit. Humans write the runbooks. Agents just run them.
SaaS with sprawling infra
10 to 100 services, queues, workers, cron jobs, and tenant-specific environments. One engineer can't hold it all.
Cloud runs it for your team
Desktop runs it on your machine
Cloud: always-on team investigations triggered by production signals.
Desktop: local, hands-on runbook creation and incident investigation.
Independent products. Neither requires the other, and they do not share accounts or investigation data.
BitSentry Cloud
For teams
BitSentry Desktop
Open source
Works with your stack
Extends to the rest of it
Every integration is a code plugin. Add one and the AI can use it in every investigation. Pull context in, push findings out.
- 19 integrations available
- Bring your own AI provider
- Runbooks run any CLI or API
- New sources added on request
Sentry
Error tracking & alerts
Wazuh
Security events & SIEM
PostHog
Product analytics & events
Trello
Incident boards & tickets
Jira
Issues from a diagnosis
ClickUp
Tasks from a diagnosis
iTop
ITSM tickets from a diagnosis
Outline
Search and write team docs
GitHub
Code & security alerts
Claude Code
Agentic investigation
Codex
AI coding agent
Gemini
Google AI models
OpenCode
Open-source agent
Cursor
AI editor agent
Groq
Fast inference
OpenRouter
Any model, one API
KiloCode
AI coding agent
Ubuntu
Runbooks over SSH
RHEL
Enterprise Linux servers
Debian
Stable server fleets
SUSE
Enterprise Linux servers
CentOS
Legacy server fleets
Bring us one recurring incident
We will help you turn it into a reusable diagnostic runbook and measure how much faster your team reaches root cause.