Trace what your agent did.
Not just what it called.
Existing LLM observability tools log API calls. iAgentLog logs agent decisions: what counterparty was chosen, why, what it cost, what the outcome was. Designed for autonomous agents from day one, not retrofitted from chatbot logging.
iAgentLog provides four primitives
Decision tracing
Capture the agent's decision tree, not only its LLM calls: counterparty selection, tool choice, rejection reasons, fallback paths.
Counterparty ledger
Every agent-to-agent interaction logged with trust signal (via iAgentFi), payment (via iAgentPay), and outcome.
Goal-drift detection
Detect when an agent's behavior deviates from its initial goal specification. Alerting before the deviation produces bad outcomes.
x402 spend analytics
Break down spend by counterparty, capability, outcome. See which integrations are paying off.
The gap iAgentLog fills
LLM-call logging misses the point
Helicone, Langfuse, LangSmith are excellent at tracing LLM calls. But an agent's value (or failure) isn't in the call. It's in the decisions between calls.
Your agent will do things you don't expect
At scale, autonomous agents produce emergent behaviors. Observability must track intent vs. action, not just execution.
Integrates with what you have
iAgentLog accepts Langfuse/Helicone/LangSmith data as input. Instead of replacing them, it adds the agent-behavior layer on top.
Planned endpoints in Shipping Q2-Q3 2026
Preview of the planned API surface. OpenAPI 3.1 specification at /.well-known/openapi.yaml. Endpoints at api.iagentlog.com will serve requests at shipping q2-q3 2026; agent-consumable JSON by design.
POST /v1/trace |
Record a decision, tool call, or outcome |
GET /v1/agent/{id}/timeline |
Retrieve timeline of decisions for an agent |
GET /v1/agent/{id}/drift |
Goal-drift analysis |
POST /v1/ingest/langfuse |
Import from Langfuse for unified view |
iAgentLog is one layer
Sixteen products. One stack. One entity. Trust, discovery, observability, payments, safety, simulation, composition, memory, identity, legal, markets, and owned compute underneath. Each layer reinforces the others. Use one or use them all.