Developer Instrumentation for AI Decisions
Add our SDK alongside your existing OpenAI, Claude, or LangChain code. We capture structured records of every decision, trace, and policy check — so you can debug, audit, and govern your AI workflows.
Execution Traces
Multi-Step Workflow Capture
Wrap any AI workflow in client.trace(). Each model call, tool use, and retrieval step is recorded as a .step() with latency, token count, and metadata.
Nested Spans
Traces support nested steps so you can see parent-child relationships across multi-agent or multi-model pipelines.
Structured Metadata
Attach custom key-value metadata to any trace or step. Filter and query by metadata fields in the dashboard.
Status & Error Tracking
Each trace and step carries a status (running, completed, failed). Surface errors with stack traces and context.
Decision Audit Trail
Structured Decision Records
Record what your AI decided, the reasoning, confidence score, and risk level using client.decision().
Action Lifecycle
Track decisions through proposed, approved, rejected, and executed states with timestamps and actor IDs.
Risk Scoring
Assign risk scores (0-1) to decisions. Filter the dashboard by risk level to focus on what matters.
Human Review Workflow
Flag high-risk decisions for human approval. Reviewers approve or reject from the console with comments.
Policy & Risk Engine
Policy Evaluations
Attach guardrail checks to decisions using client.policy_evaluation(). Record pass/fail status and evaluation details.
Risk Signal Aggregation
View all risk signals — failed policies, high-risk decisions, anomalous patterns — in a single dashboard view.
Webhook Alerts
Configure webhooks to fire when policies fail, decisions are blocked, or risk thresholds are exceeded.
Retention Policies
Set configurable data retention periods. Automate cleanup of old traces and decisions to meet compliance requirements.
Console Dashboard
Trace Explorer
Browse and search execution traces. Drill into individual steps, view timing, and inspect metadata.
Decision Feed
See all AI decisions across your systems with filters for risk level, status, and time range.
Review Queue
A dedicated view for decisions awaiting human review. Approve or reject with one click.
Compliance Exports
Export decisions, traces, and policy evaluations as CSV or JSON with configurable date ranges.
Works with Your Existing Stack
AI Logbook is not a proxy. Install the SDK, wrap your calls, and records flow to our API. No changes to your AI provider or framework.
SDKs
- Python (pip install ailogbook)
- Node.js (@ailogbook/sdk)
- REST API (any language)
AI Providers
- OpenAI
- Anthropic
- Any LLM (via SDK instrumentation)
Frameworks
- LangChain (manual instrumentation)
- CrewAI (manual instrumentation)
- AutoGen (manual instrumentation)
Infrastructure
- REST API
- Webhooks
- CSV/JSON Exports
Use Cases
Audit Trail for Regulated AI
Financial services, healthcare, and legal teams need proof of what the AI decided, who approved it, and what happened. Export structured records for auditors.
Debug Agentic Workflows
Trace multi-step agent runs end to end. Find exactly where failures, hallucinations, or unexpected tool calls happen.
Human-in-the-Loop Governance
Flag risky AI decisions for human review before execution. Build approval workflows with confidence thresholds and policy checks.
Compliance Reporting
Export structured decision and trace records as CSV or JSON for SOC 2, GDPR, and internal audit processes.