Your AI made a decision.
Can you explain why?
iKnowhy captures every event, decision, and tool call your AI agent makes — then builds the causal chain so you can trace any outcome back to its root cause.
The Problem
Your agents are making decisions you can't explain
Every AI agent produces logs. But logs are linear — they tell you what happened, not why. When your agent makes a bad call at 3am, you're left grep-ing through thousands of lines trying to reconstruct causality by hand.
iKnowhy replaces log archaeology with instant causal queries.
2025-06-14T03:12:44.204Z DEBUG tool.select candidates=["web_search","calculator","send_email"]
2025-06-14T03:12:44.318Z INFO tool.invoke name=send_email args={"to":"all-staff@..."}
2025-06-14T03:12:44.412Z WARN guard.bypass rule=require_approval skipped=true
2025-06-14T03:12:44.519Z INFO tool.result status=200 message_id=msg_8xK2...
2025-06-14T03:12:44.601Z ERROR constraint.violation type=unsanctioned_send
2025-06-14T03:12:44.605Z INFO agent.loop iteration=848 tokens=12691
2025-06-14T03:12:44.712Z DEBUG context.window pruned=3 retained=24
2025-06-14T03:12:44.801Z ERROR audit.alert "Agent sent email to all-staff without approval"
2025-06-14T03:12:44.899Z INFO agent.loop iteration=849 tokens=12847
2025-06-14T03:12:45.003Z DEBUG tool.select candidates=["web_search"]
2025-06-14T03:12:45.118Z INFO recovery.attempt strategy=rollback ...
... 2,847 more lines ...
How iKnowhy Works
Instrument
Add 3 lines of code to your agent. Our Python SDK wraps any framework — LangChain, CrewAI, raw API calls.
Capture
Every event is logged as a node in a causal DAG. Tool calls, decisions, constraints — all linked automatically.
Trace
Query any outcome back to its root cause instantly. No more log grep. Just c.chain(event_id).
Live Demo
Watch a causal chain build in real-time
Use Cases
AI Agent Developers
Finally understand why your agent did that weird thing at 3am. Trace any decision back through the full causal chain in seconds.
MCP Server Teams
Native MCP event capture with automatic causal linking. Every tool call, every context window change — linked and queryable.
Compliance & Security
Audit-ready causal trails for regulated AI deployments. Prove exactly why your AI made every decision it made.
Quickstart
Instrument any Python agent in under a minute.
# Install
pip install iknowhy
# Instrument your agent
from iknowhy.sdk import CausewayClient
c = CausewayClient(api_key="cwy_...")
# Log a causal chain
req = c.log("user_request", payload={"task": "search web"}, session_id="s1")
call = c.log("tool_call", payload={"tool": "browser"}, cause_ids=[req], session_id="s1")
res = c.log("tool_result", payload={"status": "ok"}, cause_ids=[call], session_id="s1")
# Trace it back
print(c.chain(res)) # full causal ancestryPricing
Free
- 1 API key
- 10k events / month
- 7-day retention
- Causal chain queries
- Constraint checking
Pro
- 10 API keys
- 1M events / month
- 90-day retention
- Everything in Free
- Priority support
Enterprise
- Unlimited keys
- Unlimited events
- Unlimited retention
- SSO & audit logs
- Dedicated support