Impraint Guard
Governance, observability and compliance reporting for your AI agents. See what they do, set limits they cannot cross, and keep a record you can show.
DPDP and OWASP Agentic reports today · more frameworks later

Integrate Guard into your agent stack
Installation guides, SDK reference, and integration walkthroughs for Anthropic, Gemini, LangChain, and LangGraph.
What Guard does
Everything you need to run agents with confidence
Observe
See every step
Tokens, cost and model for every step, with agent-to-agent escalations linked together.
Tracetr_8f2a · support-agent
CompletedRetrieve
orders-index · 3 documents
Model call
gemini-2.5-flash · 1,820 tokens
Escalated to refund-agent
Linked trace
Observe
Pause any agent
See each agent's activity and stop one with a click.
Agents3 agents
support-agent
1,284 runs today · $3.42
refund-agent
212 runs today · $0.88
onboarding-agent
Paused by admin · 10:42
Observe
Catch what is unusual
Per-agent baselines flag cost and behaviour spikes.
Anomaliescost per run
Last 7 daysCost spike detected
Alert sent to Slack and email
Protect
Find sensitive data and attacks
Scan inputs and outputs for PII and prompt injection, including Indian ID formats.
Content safetyinput · onboarding-agent
2 findingsPlease update my KYC. My PAN is ABCDE1234F and Aadhaar is 2345 6789 0123.
PAN number
Indian ID pack · detected in input
Aadhaar number
Indian ID pack · detected in input
Protect
Set limits that hold
Block, throttle or hold a call when a rule is broken.
Rulessupport-agent
New ruleCost per hour exceeds $5
When this happens, Guard will:
Circuit open
2 calls blocked · resets in 14 min
Protect
Keep a human in the loop
Approve or reject risky actions from Slack.
Slack#guard-approvals
Waiting 02:14Impraint Guard
APP · 10:42
Approval needed
refund-agent wants to issue a refund of ₹48,000 on order #1042. This is above the approval limit.
The agent waits for a decision before continuing.
Respond
Hear about it first
Slack and email, without repeat alerts.
Alertslast 24 hours
3 deliveredCost spike
support-agent · 4.2× baseline
PII detected
onboarding-agent · PAN and Aadhaar
Approval requested
refund-agent · above limit
Report
Show your work
DPDP and OWASP Agentic reports as PDFs, backed by a tamper-evident audit log.
Compliance reports
PDFDPDP Act report
Generated Nov 4 · all agents
OWASP Agentic AI report
Generated Nov 4 · all agents
Audit log verified
Tamper-evident hash chain · 12,480 entries