AI decisions you can explain, audit, and reuse.
Every decision deserves a why.
The decision tracing and governance layer for AI agents. Works with LangChain, n8n, and any custom agent—no orchestration required.
from aiagentree import AgentreeClient, TracedAgent
# Auto-trace any LLM (3 lines of code)
with TracedAgent(client, workflow_id="claim_review", entity_id="CLM-4821") as agent:
response = agent.chat(
llm=openai_client, # or anthropic, ollama, langchain
message="Review this insurance claim",
context={"claim_amount": 15000, "customer_tier": "premium"}
)
# Reasoning automatically extracted & traced!
print(f"Trace ID: {agent.trace_id}")
AI agents are making critical decisions, and no one can explain why.
Your AI agents work 24/7, handling customer refunds, reviewing contracts, and approving payments. But when something goes wrong—or when regulators ask "why?"—you're stuck.
Black Box Problem
AI models produce outputs, not explanations. The reasoning disappears into logs and prompts.
Trust Gap
Legal doesn't approve what they can't audit. Customers don't accept what they can't appeal.
Improvement Gap
You can't systematically improve what you can't understand. Optimization becomes guesswork.
Scale Problem
Manual oversight worked for one agent. It doesn't work for thousands of decisions per hour.
They store what. We store why.
Every tool tells you what happened. No one tells you why it happened.
Until now.
WHAT
What everyone else stores
- The agent ran these steps
- It took 2.3 seconds
- It used 1,500 tokens
- The output was "approved"
"What did it do?"
Useful. But incomplete.
WHY
What we capture
- The decision was to approve because...
- These policies were evaluated...
- This exception was allowed when...
- Similar decisions succeeded 94% of the time
"Why did it decide that?"
The question that matters.
No one owns "why" today.
Execution tools log behavior. Observability tools track metrics.
We capture the reasoning that makes decisions defensible.
Decisions that compound.
Every outcome makes the next decision better. Bad precedents automatically sink; proven ones rise.
Everything You Need for AI Decision Governance
From decision capture to pattern detection, we provide the complete infrastructure for responsible AI operations.
Decision Trace Capture
Capture the complete reasoning behind every AI decision—from inputs to deliberation to final output. Structured data, not unstructured logs.
Validation & Quality Scoring
Automatically validate that AI decisions meet quality standards. Get a quality score (0-100) for every decision and catch issues early.
Precedent Search
Find similar past decisions using vector embeddings. AI-powered institutional memory that learns from every decision.
Outcome Tracking
Connect AI decisions to real-world results. Track outcomes at three horizons: immediate, short-term, and long-term.
Pattern Detection
Identify recurring decision patterns that could be automated. Let AI suggest its own improvements based on historical data.
Approval Workflows
Human-in-the-loop when you need it, with SLA tracking to ensure timely review. No decisions stuck in limbo.
Real-Time Monitoring
See AI decisions as they happen. Live dashboards, instant alerts, and never be surprised by what your AI is doing.
Analytics & Reporting
Comprehensive analytics on AI decision-making. Executive dashboards, agent performance, and audit-ready exports.
Built for Real-World AI Operations
See how organizations across industries use Decision Agent Platform to govern their AI systems.
Customer Service AI
All IndustriesTransparent AI-powered customer interactions with complete decision tracing for trust and compliance.
Dramatic reduction in decision explanation time
Contract Review
LegalAI-powered contract analysis with transparent risk scoring and precedent-based recommendations.
Significantly faster review cycles
Insurance Claims
InsuranceCompliant claims processing with complete audit trails for every approval, denial, and escalation.
Faster appeal processing with full context
Financial Operations
FinanceAudit-ready financial automation with complete decision documentation for every transaction.
Audit-ready decision documentation
HR & Recruiting
Human ResourcesFair, transparent hiring with documented evaluation criteria and defensible decision trails.
Defensible hiring documentation
IT Operations
TechnologyIntelligent IT automation with decision traces for incident response and change management.
Faster post-incident analysis
Get Started in Minutes
Add decision tracing to your AI agents with just a few lines of code.
Integrate SDK
Add our SDK to your agent code. Works with LangChain, n8n, or any custom framework. No orchestration changes required.
pip install ai-agentree-sdkTrace Decisions
As your agent runs, the SDK captures structured decision events—inputs, deliberation, relations, and outcomes.
- Policy checks
- Exceptions & approvals
- Justifications
Analyze & Improve
View decisions in real-time, search precedents, track outcomes, and identify patterns for automation.
- Live dashboard
- Pattern detection
- Audit reports
Six Ways to Connect Your AI Agents
From native SDKs to industry-standard protocols. Choose the integration method that fits your architecture—or use them all.
Python SDK
Native IntegrationFull-featured Python SDK with auto-retry, PII redaction, and structured output parsing. Three lines to start tracing decisions.
pip install ai-agentree-sdkREST API
Universal AccessFull REST API for any language or framework. Trace ingestion, search, validation, analytics—everything the SDK does, via HTTP.
POST /api/aiagentree/tracesMCP Protocol
AI-NativeModel Context Protocol support via Streamable HTTP. AI agents discover and call decision tools natively—no SDK installation required.
4 tools: append, seal, get, searchAgent-to-Agent
Multi-Agent DelegationDelegate decisions between AI agents with full trace lineage. Parent-child chains up to 10 levels deep, with complete Decision Packets at each hop.
POST /api/aiagentree/a2a/delegateOpenTelemetry
Observability BridgeAutomatically emit decision spans to your existing observability stack. Works with Jaeger, Datadog, Grafana Tempo, and any OTLP-compatible backend.
OTEL_ENABLED=true # zero-code setupSchema Export
Auto-DiscoveryAuto-generated JSON Schema and OpenAPI 3.1 specs for every data model. Use them for validation, code generation, or documentation.
GET /api/aiagentree/schemas/openapiChoose Your Path. Or Use Them All.
Every integration method talks to the same decision engine. Mix and match based on your architecture—SDK for your Python agents, MCP for Claude and ChatGPT, A2A for multi-agent workflows, OTel for your ops team.
Build Fast
Zero-config auto-tracing extracts reasoning from any LLM—OpenAI, Anthropic, Ollama, LangChain. Add decision tracing in 3 lines of code.
Connect Natively
MCP and A2A let AI agents discover and use decision tools without any custom code. True AI-native integration.
Observe Everything
OpenTelemetry bridges decision spans into Jaeger, Datadog, or Grafana—see AI reasoning alongside your application traces.
What AI Agentree is NOT
We complement your existing tools—we don't replace them.
Not an agent framework
We don't orchestrate agents. Use LangChain, n8n, or your own code.
Not a prompt manager
We don't manage prompts. We capture decisions that result from any prompt.
Not a chain-of-thought viewer
We capture structured justifications, not raw model outputs.
The decision layer those tools don't provide
We record decisions as first-class artifacts for audit, learning, and improvement.
Eenvoudige, transparante prijzen
Begin gratis, schaal naarmate u groeit. Alle plannen bevatten kernfuncties.
Free
For development and proof-of-concept
- 25 traces/month
- Core tracing features
- 7-day data retention
- Community support
Basic
For your first production agent
- 5,000 traces/month
- All features included
- 30-day data retention
- Email support
Team
For growing teams
- 5,000 traces/month
- 60-day data retention
- 5 team members
- Priority support
Professional
For teams with multiple AI agents
- 15,000 traces/month
- 90-day data retention
- API access
- Custom dashboards
Enterprise
Voor grootschalige AI-activiteiten
Frequently Asked Questions
Looking for Human Discussion Tools?
Check out Argumentree Forum for team debates, consensus building, and decision visualization.
Every decision deserves a why.
Give your AI decisions the answer they deserve.
When someone asks "why did the AI do that?"—you'll have the answer. Not a log file. Not a guess. The actual reasoning.