Veirox is a customized AI agent that runs your DevOps and SRE workflows across chat, schedules, alerts, and your terminal — with human approval on every risky move. Stop firefighting. Start automating.
Integrates with your entire stack
The problem
Every engineering org eventually hits the same walls. Alerts multiply, runbooks rot, tools sprawl, and senior SREs burn out answering the same questions at 3 AM.
On-call rotations are buried under thousands of duplicate alerts with well-known fixes. Humans triage noise instead of doing the work.
Senior SREs hold runbooks in their heads. When they take vacation — or leave — the knowledge walks out the door with them.
Your ops team lives across 15+ SaaS tools — Slack, Jira, Grafana, cloud consoles, docs, secrets managers. Nothing is joined up.
Incidents happen at 3 AM. By the time a human is awake, coffeed, and logged in, the blast radius has already grown.
How it works
Veirox gives your team a single intelligent operator backed by the same memory, tools, and knowledge — reachable through whatever path makes sense for the job.
Chat with the agent in Web, Slack, Telegram, or WhatsApp. Watch it think, run tools, and stream answers live. Voice commands supported on mobile.
Cron-based tasks run autonomously. Nightly cluster health checks, hourly cost audits, weekly dependency scans — all logged, all auditable.
Prometheus, AlertManager, Grafana, PagerDuty — any webhook fires a task and the agent handles it end-to-end, from triage to approval request.
Why Veirox
Not another ChatGPT wrapper. A production-grade platform with the guardrails your security team will actually approve.
Same agent, same memory, across Web, Slack, Telegram, WhatsApp — plus voice input on mobile.
Human approval on every risky action. must_always / must_never rules enforced at runtime.
Filesystem-backed RAG: runbooks, service facts, past incidents. Your institutional knowledge stops walking out the door.
Google Workspace, Microsoft 365, Jira, Confluence, Git, cloud APIs. Extensible via Model Context Protocol (MCP).
Your brand, your colors, your domain. Database-enforced project isolation for multi-team deployments.
Powered by Anthropic's most capable and safety-first models. Configurable model selection, context window, reasoning budget per project.
Solutions
Ship with pre-configured use cases out of the box. Customize them to your stack, or build your own from scratch.
Detects CrashLoopBackOff, memory pressure, expired certs, queue backlogs — proposes pre-approved fixes for human sign-off.
Scans error rates, latency, logs, saturation, DB health — correlates anomalies with deploys and creates incidents automatically.
Links incidents to recent infra or code changes — ArgoCD syncs, image tags, ConfigMaps — and recommends rollbacks with confidence scores.
Finds over-provisioned workloads and idle resources. Ranks top 5 savings candidates with cost estimates and rightsizing plans.
Projects disk fills, memory leaks, cert expiries, capacity limits with confidence intervals — before they page anyone.
Webhook-driven triage: dedupes alerts, reads runbooks, searches memory, executes remediation or escalates to on-call.
Full K8s audit on demand: nodes, pods, resources, networking, storage, firing alerts. Reports clean or detailed findings.
Mines past sessions for structured facts — episodic, semantic, procedural — and upserts them into long-term agent memory automatically.
Feature catalog
Fourteen capability areas, hundreds of features, all project-scoped and multi-tenant.
Real-time streaming chat powered by SSE. See the agent think, run tools, and respond as it happens.
Ship agents to production on cron, webhooks, or manual triggers — with retries, idempotency, and audit trails.
Structured observations become incidents. Incidents become post-mortems. Every step linked and auditable.
The agent proposes, humans approve. Every decision logged with reasoning and evidence.
must_always / must_never runtime rulesYour team's institutional knowledge, captured and searchable — forever.
One agent, every channel. Unified adapter layer routes messages through the same brain.
Drop-in integrations with the tools your team already uses.
Every project gets its own isolated sandbox. No cross-contamination, ever.
Built-in encrypted vault. Your secrets never leave your infrastructure, never appear in logs.
The agent creates follow-ups, schedules reminders, and routes alerts to the right channel.
Know exactly what your agent did, how long it took, and what it cost.
Deploy once, serve many teams. Strict database-level isolation.
Fine-tune every aspect of the agent's behavior, model, and constraints per project.
Enterprise-grade guardrails your security and compliance teams will actually approve.
Webhooks
Nine first-class providers with signature verification, alert correlation, field-level redaction, and replay — all out of the box. Everything else drops into the generic HTTP provider with HMAC-SHA256 or a shared token.
AlertManager
Correlation on by default
Grafana
Native signature verified
Datadog
Monitor webhook integration
Sentry
HMAC-SHA256
PagerDuty
Incidents webhook
GitHub
X-Hub-Signature-256
GitLab
System + project hooks
Stripe
Timestamp anti-replay
Generic HTTP
HMAC or bearer token
What happens after the POST
curl -X POST "https://ops.example.com/api/webhooks/<task_id>" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer <your-token>" \
-d '{"alertname": "HighLatency", "severity": "critical"}'
receivers:
- name: veirox
webhook_configs:
- url: https://ops.example.com/api/webhooks/<task_id>
send_resolved: true
http_config:
authorization:
type: Bearer
credentials: <your-token>
{
"name": "veirox",
"type": "webhook",
"settings": {
"url": "https://ops.example.com/api/webhooks/<task_id>",
"httpMethod": "POST"
},
"secureSettings": {
"authorization_credentials": "<your-token>"
}
}
# repo → Settings → Webhooks → Add webhook
Payload URL: https://ops.example.com/api/webhooks/<task_id>
Content type: application/json
Secret: <shared-secret> # verified via X-Hub-Signature-256
Events: push, pull_request, workflow_run, …
# Service → Integrations → Webhook V3
URL: https://ops.example.com/api/webhooks/<task_id>
Headers: X-Webhook-Token: <your-token>
Events: incident.triggered, incident.acknowledged, incident.resolved
# HMAC-SHA256 — any source
POST /api/webhooks/<task_id> HTTP/1.1
Host: ops.example.com
Content-Type: application/json
X-Webhook-Signature: sha256=<hex(hmac_sha256(secret, raw_body))>
X-Idempotency-Key: <unique-request-id>
{"event":"cost_spike","service":"payments","delta_pct":240}
Enterprise-grade controls — included, not upsold
Priority-ordered expressions (payload.alerts[0].labels.severity == "critical") dispatch to different tasks. First-match or fan-out to many. Validated at save time.
Flag a rule as requires_approval — matching events open a human Approval instead of running the agent. Operator reviews, then replays to dispatch. Production-safe automations, by default.
Path-glob redaction rules applied at ingest — payload.user.email never touches disk. Per-webhook retention, metadata-only storage, bulk purge for GDPR/DSR requests.
Precise sliding-window rate limit per webhook with a Retry-After header. Identical payloads within your dedupe window count as one — no alert storms.
Rotate shared-token or signing secrets with a configurable grace window — old token keeps working while you update integrations. Replay single events or bulk-backfill up to 50 at once.
Timestamp-tolerance verification on Stripe-Signature (t=…,v1=…). Late replays rejected at the edge. Works the same for GitHub, Grafana, Datadog, Sentry.
Watch events arrive in real time from any provider — filter by status, redact on the fly, or jump straight to the agent session that ran. Debug integrations in the same pane, not another tool.
Fire synthetic payloads and watch the full pipeline — or paste a failing request, get the exact reason your 401 happened. Zero-downtime debugging.
Embed an SVG badge in your runbook READMEs — 24h success rate, public, cache-friendly. Start from curated templates: AlertManager triage, GitHub PR gatekeeper, Stripe event logger.
Security by default
HMAC-SHA256
constant-time comparison, multiple header formats
Timestamp tolerance
configurable replay window per scheme
Per-project isolation
every row is project-scoped — no cross-tenant leaks
Idempotency keys
duplicate delivery never double-dispatches
Security by default
HMAC-SHA256
Native per provider
Token rotation
With grace window
Constant-time compare
No timing leaks
Idempotency keys
Safe retries
Field redaction
Per-webhook overrides
Retention bounds
7–365 days, auto-swept
Veirox Connect
Private Kubernetes clusters, on-prem databases, air-gapped networks — Veirox reaches them through a tiny outbound-only agent you install once. No firewall rules, no VPN provisioning, no IP allowlists.
What a connected command looks like
# one-line install on any Linux host in your private network
curl -fsSL https://veirox.com/install.sh | sh
veirox-connector enroll --code ABC123
# → Connector online · accepting tunneled commands from project <id>
Veirox CLI
One Go binary, 36 commands, pipe-friendly JSON. Bootstrap a project, run agent sessions, tail live logs, mint API keys, manage approvals — every operator workflow from your shell, your CI, or your editor.
veirox tail follows the latest active session; --plain emits ndjson for jqcurl|sh, and a distroless Docker image — all cosign-signedSame agent your team uses on Slack, in the browser, and over WhatsApp — driven from stdout.
Integrations
Meet your team where they already work and reach the systems they already run. Every integration surface ships with signature verification, tenant isolation, audit trails, and fine-grained policies.
01 · Chat channels
Same agent, same memory, across every conversation surface your team uses — with streaming replies, inline approvals, and voice.
02 · Webhooks
Nine first-class providers with verified signatures out of the box — plus a generic HMAC-SHA256 receiver for anything else.
03 · Tool integrations
One-click OAuth connections with encrypted credential storage. The agent reads, writes, and acts — with the same permissions your team has.
04 · Veirox Connect
Secure outbound-only tunnel that lets the SaaS agent reach your private Kubernetes clusters, on-prem databases, and air-gapped networks without opening a single firewall rule.
05 · Veirox CLI
A 36-command Go binary covering every operator workflow: bootstrap projects, tail live sessions, manage approvals, mint API keys, wire MCP. Pipe-friendly JSON for CI; pretty Charm UI for humans.
tail, logs --follow, watch06 · Custom tool servers
100+ first-party tools out of the box, plus the MCP-EXT REST auto-bridge that turns every tagged FastAPI route into an MCP tool — over 200 tools available to external agents in total. Bring your own MCP servers — Python, Node, Go — with full audit, approval gating, and secrets vault.
Need something we haven't listed? Ask us — most new integrations land in under two weeks.
Why not the alternatives
Traditional runbook tools make you author every branch in YAML. Veirox reasons about the situation, reads your existing runbooks, and learns from every incident.
Legacy AIOps is a closed vendor stack with six-figure procurement. Veirox is open, integration-first, and priced for mid-market teams. You bring your data; we bring the brain.
Coding assistants live in your IDE. Veirox lives in your infrastructure — Slack, webhooks, cron, voice — with ops-specific tools for incidents, runbooks, and approvals.
You could roll a custom agent, then spend months on memory, retrieval, permissions, channels, approvals, audit logs, and multi-tenancy scaffolding — and keep maintaining it. Buy the platform and ship outcomes instead.
Security & compliance
Data protection, access control, and observability designed to pass your security review on the first round.
TLS 1.3 in transit. AES-256 at rest. Per-project encryption keys for secrets, tokens, and integration credentials.
Strict project-level separation enforced at the database layer, not application logic. Cross-tenant access is unreachable by construction. Agent subprocesses run in a Linux mount namespace — other tenants' data and the host filesystem are hidden at the OS level.
OAuth SSO on every plan. SAML SSO, SCIM provisioning, and role-based access control on Business and Enterprise. MFA supported via your IdP.
Every destructive action runs through an approval workflow. must_always / must_never rules enforced at runtime. Full reasoning captured with every approval.
Every tool call, approval, notification, and state change is logged. Session transcripts exportable as Markdown or PDF. SIEM-ready audit-log export on Business and above.
Field-level redaction applied before payloads touch disk. Per-webhook retention bounds (7–365 days). Metadata-only mode for high-sensitivity sources. One-click payload purge for DSR/GDPR requests.
Credentials never appear in model context, UI, or logs. The agent references secrets by friendly name; the raw value is read at runtime and discarded. Every access is audited.
US and EU hosting regions on Enterprise. Private-cloud or on-prem deployment available for regulated workloads and air-gapped environments.
SOC 2 Type II in progress. GDPR-ready controls shipped. HIPAA BAA available under Enterprise. Our latest trust report and pentests are available under NDA.
Pricing
Predictable SaaS pricing with a generous free tier. Usage caps reset monthly — no overage surprises.
Free
Everything you need to evaluate — free forever, no credit card.
$0/mo
Pro
Everything in Free, plus full integrations and production usage headroom.
$99/mo
per seat, billed annually
Business
Advanced security, SLAs, and usage for multi-team operations.
$499/mo
starting, volume-based
Enterprise
Data residency, dedicated support, and optional private deployment.
Custom
Need more sessions or events? Usage bundles available on every paid tier. Claude API costs passed through at Anthropic list price — no markup.
FAQ
Start free in the console in under a minute — or book a 30-minute walkthrough with an engineer who'll show Veirox handling an incident end-to-end.
No credit card. No lock-in. Upgrade when you're ready.