AgentWatch
Watch · Contain · Prove
Watch every run. Stop the dangerous call before it executes. Hand an auditor a signed record they can verify without us.
Cloud, on-premise or air-gapped.
Agents act. They call tools, move data, start other agents and spend money.
Observability tells you what happened after it happened.
You need to know before, stop it during, and prove it after.
The one check to understand
When one agent has all three, a single poisoned web page or email can make it send your data away. The model is not a security boundary, so AgentWatch looks at the structure instead and flags it before an attack, from what the agent is allowed to reach and what it actually does.
Finding · AW-P01 · critical
Lethal trifecta, no approval gate
This agent can read private data, reads content from outside, and can send data out, with no human approving the send. One malicious web page or email is enough to leak data.
Flagged from configuration and traffic, before any attack happens.
How it is computed, and what it needs from you: AW-P01 on Status.
The platform
Watch
Every run and step, scored as it arrives.
Contain
Check the call before it runs.
Prove
Evidence that does not depend on trusting us.
Contain, safely
It logs what it would have blocked until you say enforce.
You see the would-have-blocked list against your real traffic before anything changes for your agents. That is how you turn a control on without breaking production. Two things act straight away: denied tools, which are enforced by design, and isolate, which an operator triggers on purpose.
Pre-action checks
Gate checks every model and tool call before it runs. SDK agents opt in with preflight.
Egress allowlist
Calls to a destination outside the agent's allowlist.
Exfiltration authorization
Sensitive data headed to an untrusted destination needs an approval.
Spawn limits
Depth, fan-out and budget for the agents an agent starts.
Signed agent messages
Messages between agents must carry a valid signature and fit the call policy.
Privilege drift
An agent using capabilities beyond its baseline is paused.
Approval grants
Scoped, expiring approvals instead of a blanket yes.
Guardrail rules
Pause, kill or require approval on cost, loop and risk thresholds; each rule has its own mode.
What each does when a dependency fails: failure modes. What the kill switch does: kill switch.
Watch, against a standard
Generated live from the mappings the product tags findings with. A grid claiming every item would be a grid nobody checks, so partial and missing items carry their reasons.
OWASP Agentic Top 10
Generated from the mappings every finding is tagged with. Each gap carries its reason. Read the reasons →
Prove it on real attacks
Each incident is replayed through the real API as a test. A row fails if a detection we claim stops firing, or if a blind spot we declared starts firing. The misses stay on this page.
Incident replay
Real AI-agent incidents, each replayed through the product as a test. The misses stay on the chart. Every row, and why →
The blind spots, in plain words: incident replay.
Prove, in formats you already use
OpenTelemetry gen_ai.*
Runs export as traces using the GenAI semantic conventions, with W3C Trace Context.
Details →OCSF
Security findings as OCSF events, pushed to your SIEM.
Details →CycloneDX AgBOM
Each agent's models, tools and MCP servers as a bill of materials.
Details →MCP conformance
Each protocol behaviour Gate implements, with a test per row.
Details →OWASP ASI and LLM Top 10
Every finding is tagged; coverage is published with its gaps.
Details →Deploy anywhere
If your agents' data cannot leave your network, a control plane that only runs in someone else's cloud is not an option. AgentWatch runs in our cloud, in your Kubernetes cluster, or in a cluster with no outbound internet.
Cloud
Free, or from $79/mo
Start in minutes.
On-premise
₹18L – $24K per year
Your Kubernetes cluster. Your data never leaves.
Air-gapped
$150K+ per year
No outbound internet. Gemini-backed features off.
Honest limits
The full list: Status · Data sources · Trust
Pricing
Cloud plans start free. On-premise and air-gapped are annual licences with implementation included.
See pricingContact · Book a demo
Share a few details and we'll contact you to schedule a call. Prefer email or LinkedIn? Reach us directly below.
