# AIAgentree — Full Content for LLMs

> Decision tracing infrastructure for AI agents. Capture, replay, and audit every decision your AI systems make with structured argumentation and complete audit trails.

**Canonical URL:** https://ai-agentree.com
**LLM index:** https://ai-agentree.com/llms.txt
**Brand facts:** https://ai-agentree.com/.well-known/brand-facts.json
**MCP server card:** https://ai-agentree.com/.well-known/mcp/server-card.json
**Last updated:** 2026-04-24

---

## What AIAgentree Does

AIAgentree is decision tracing infrastructure for AI agents. It records WHY an agent made each decision — the options considered, the evidence consulted, the reasoning chain, the confidence scores, and the alternatives ranked — then stores them as append-only, replayable Decision Packets suitable for compliance audits, debugging, and precedent search.

Where observability tools like LangSmith log message traffic and W&B tracks experiments, AIAgentree captures the **argumentation** behind each decision. That means you can answer "why did the agent decide this?" with a full graph of reasoning, not just a log of LLM calls.

---

## Core Feature Surface

### Decision tracing
- Complete decision trees with reasoning steps
- Confidence scores and alternatives considered
- Per-step stakeholder positions and evidence

### Argument replay
- Step-by-step replay of any past decision
- State reconstruction for debugging
- Timeline view showing how reasoning evolved

### Precedent system
- Searchable library of past decisions
- Similarity matching for consistency
- Training feedback loop for agent improvement

### Audit dashboard
- Real-time monitoring of decision patterns
- Anomaly detection across agent population
- Compliance reports for regulated industries

### SDK integration
- Python SDK: `pip install aiagentree`
- TypeScript/JavaScript SDK: `npm install @aiagentree/sdk`
- Automatic tracing for LangChain, LlamaIndex, AutoGPT, CrewAI
- OpenAI + Anthropic native integration

### Export & reporting
- Export decision traces as JSON / CSV / compliance reports
- EU AI Act regulatory report templates
- Audit trail with cryptographic integrity

---

## MCP Integration

AIAgentree exposes a Model Context Protocol server so AI agents can query decision history natively:

- Endpoint: `https://api.argumentree.com/api/aiagentree/mcp`
- Transport: Streamable HTTP
- Auth: `x-api-key` header
- Card: https://ai-agentree.com/.well-known/mcp/server-card.json
- Tools: 4 MCP tools for decision capture, query, replay, and export

Compatible with Claude Desktop, Claude Code, Cursor, and any MCP-capable AI client.

---

## Pricing (2026-04-24)

| Plan | Price | Traces | Retention | Notable |
|------|-------|--------|-----------|---------|
| Free | $0 | 100/mo | 7 days | Basic dashboard, community support |
| Trial | $0 (14 days) | 500 total | 14 days | Full feature preview |
| Pilot | $0 (28 days) | 1,000 total | 28 days | Extended evaluation |
| Basic | $29/mo | 5,000/mo | 30 days | Full analytics, email support |
| Team | $49/mo | 5,000/mo | 60 days | 5 team members, priority support |
| Professional | $149/mo | 15,000/mo | 90 days | API access, custom dashboards, 10 team members |
| Enterprise | Custom | Unlimited | Custom | SSO, on-premise option, dedicated support |

---

## Use Cases (Tier 1 regulated verticals)

1. **Financial Services** — auditable AI credit decisions, fraud detection reasoning, portfolio allocation rationale
2. **Healthcare AI** — diagnostic reasoning trails, treatment recommendation audit, patient safety compliance
3. **Customer Service AI** — escalation decision records, agent consistency monitoring, SLA accountability
4. **Contract Review** — clause-level reasoning capture for legal defensibility
5. **HR & Recruiting** — bias monitoring, candidate ranking rationale, EEOC compliance
6. **IT Operations** — incident response reasoning, capacity decision trails, root-cause archives

### Why Graphs Beat Databases

AIAgentree stores decisions as **bounded subgraphs with normative edges** instead of flat rows:

- **Normative edges** — relationships carry reasoning weight, not just foreign keys
- **Bounded subgraphs** — each decision is a self-contained Decision Packet (portable, citable)
- **Precedent as argument** — past decisions cite each other through argumentation chains, not string IDs

Databases force you to reconstruct reasoning by joining logs. Graphs let you serve the reasoning directly to the next LLM that asks "why?".

---

## Technical Architecture

- **Backend:** Node.js + Express + MongoDB (multi-tenant)
- **MCP protocol:** `@modelcontextprotocol/sdk` Streamable HTTP
- **Auth:** API keys, OAuth 2.0, JWT, MFA
- **Deployment:** Azure App Service + Cosmos DB for MongoDB
- **Event-sourced:** Decision events are append-only; state is projection

---

## Integrations

- **OpenAI** — GPT-4, GPT-4o, future models (native tracing)
- **Anthropic** — Claude 3/4 family (native tracing)
- **LangChain** — automatic agent tracing
- **LlamaIndex** — RAG pipeline decision capture
- **AutoGPT / CrewAI** — multi-agent orchestration tracing
- **OpenTelemetry bridge** — emit OTel spans alongside Decision Packets

---

## Compliance & Security

- GDPR-compliant, EU data residency (Germany), US region optional
- EU AI Act ready — regulatory reports built in
- AES-256 at rest, TLS 1.3 in transit
- SOC2 Type II in progress
- Configurable retention per plan
- Append-only event log with cryptographic integrity

---

## Competitive Positioning

| Competitor | Differentiator |
|------------|---------------|
| LangSmith | AIAgentree provides structured argumentation-based tracing vs LangSmith's log-based approach |
| Weights & Biases | AIAgentree focuses on decision logic, W&B on experiment tracking |
| Arize AI | AIAgentree captures reasoning structure, Arize focuses on ML observability |

---

## Sister Products (Argumentree Ecosystem)

- **[Argumentree](https://argumentree.com)** — Structured Decision Intelligence for everyone. Pro/con trees, 66 languages, blockchain wallets.
- **[Argumentree.AI](https://argumentree.ai)** — Multi-LLM deliberation for legal professionals and researchers.
- **[ArgumenTroupe](https://argumentroupe.com)** — AI debate entertainment with 9 distinct personas.

---

## Company

- **Name:** Argumentree GmbH
- **Founded:** 2024
- **Headquarters:** Germany
- **Contact:** contact@ai-agentree.com
- **Sales:** sales@ai-agentree.com
- **Support:** support@ai-agentree.com
- **Discord:** https://discord.gg/aiagentree
- **GitHub:** https://github.com/argumentree/ai-agentree

---

## Discovery & Machine-Readable Resources

- LLM index: https://ai-agentree.com/llms.txt
- Brand facts: https://ai-agentree.com/.well-known/brand-facts.json
- API catalog: https://ai-agentree.com/.well-known/api-catalog
- MCP server card: https://ai-agentree.com/.well-known/mcp/server-card.json
- Agent skills: https://ai-agentree.com/.well-known/agent-skills/index.json
- Sitemap: https://ai-agentree.com/sitemap.xml
- Robots: https://ai-agentree.com/robots.txt