Open Infrastructure

Core data-model schemas (Zod) for mandates, actions, and delegations, the foundation of the Truss Action Protocol.

Cross-org evidence coordination via MCP server, connect agents across organizational boundaries.

TypeScript SDK for Truss, Ed25519 signing and TAP protocol client for Node.js agents.

Python SDK for Truss, wrap LangGraph, AutoGen, CrewAI actions with TAP provenance.

CLI tool for Truss, verify evidence, manage identities, inspect packages from the terminal.

LangGraph agent loop recording, automatically capture every node execution with TAP provenance.

AutoGen multi-agent conversation recording, trace every agent-to-agent message as evidence.

CrewAI task execution recording, capture crew workflows with signed action receipts.

LlamaIndex query and retrieval recording, log every RAG response as verifiable evidence.

OpenAI Agents SDK integration, record handoffs, function calls, and guardrail evaluations.

Semantic Kernel function invocation recording, log every AI function call with full provenance.

CVE-style registry for tracking, reproducing, and preventing AI hallucination incidents in high-stakes legal domains.

Scaffolding toolkit for building safe, jurisdiction-aware Model Context Protocol (MCP) servers for legal use cases.

Multi-source citation verification checks whether legal citations exist across CourtListener, Harvard CAP, Google Scholar, and local corpora.

Vendor-neutral JSON Schema for representing structured contract data, parties, clauses, obligations, risks, dates, financials.

Document intelligence pipeline to extract, chunk, embed, search, and ask over any document format in one command.

African Legal AI Benchmark, evaluating LLM legal reasoning across African jurisdictions.